CN108805190A

CN108805190A - A kind of image processing method and device

Info

Publication number: CN108805190A
Application number: CN201810538373.2A
Authority: CN
Inventors: 刘楠
Original assignee: Beijing QIYI Century Science and Technology Co Ltd
Current assignee: Beijing QIYI Century Science and Technology Co Ltd
Priority date: 2018-05-30
Filing date: 2018-05-30
Publication date: 2018-11-13
Anticipated expiration: 2038-05-30
Also published as: CN108805190B

Abstract

The present invention provides a kind of image processing method and devices, belong to field of computer technology.Wherein, terminal can first judge whether target image is picture mosaic image, then, it can be when target image be picture mosaic image, target image is split, obtain at least two target subgraphs that target image includes, finally, classify respectively to every target subgraph, and determine the classification of target image based on the classification results of every target subgraph, in this way, by the way that target image is split, the content complexity of target image is reduced, the accuracy rate of judgement can be improved to avoid due to the complicated caused erroneous judgement of picture material.

Description

A kind of image processing method and device

Technical field

The invention belongs to field of computer technology, more particularly to a kind of image processing method and device.

Background technology

With the continuous development of field of computer technology, the information in internet is more and more.In order to ensure in internet The safety of information, legitimacy, it usually needs the information uploaded to user is audited, could be by these when audit passes through Information is distributed in internet.For example, for the image in internet, it usually needs classify to image, to judge image Whether it is illegal image, for example, judging whether image is pornographic image or bloody violence image, etc..

In the prior art, it is typically directly treated using preset disaggregated model using pending image as basis for estimation Audit image is classified.But when pending image is the picture mosaic image of the independent subgraph composition of at least two contents, Since the composition and content of picture mosaic image are complex, the difficulty of judgement is larger, therefore the problem of often will appear erroneous judgement, sentences Disconnected accuracy rate is relatively low.

Invention content

A kind of image processing method of present invention offer and device, to solve the problems, such as that the accuracy rate judged is relatively low.

According to the present invention in a first aspect, provide a kind of image processing method, be applied to terminal, this method can wrap It includes：

Judge whether target image is picture mosaic image；

If the target image is picture mosaic image, the target image is split, the target image packet is obtained At least two target subgraphs included；

Classify respectively to every target subgraph, and the mesh is determined based on the classification results of every target subgraph The classification of logo image.

Optionally, described to judge the step of whether target image is picture mosaic image, including：

The target image is converted into gray level image, obtaining can processing target image；

Based on preset horizontal edge operator, determine described in can processing target image first edge figure in the horizontal direction, And be based on preset vertical edge operator, determine described in can processing target image vertical direction second edge figure；

The divisible quantity of the target image in the horizontal direction is determined based on the first edge figure, obtains first point Quantity is cut, the divisible quantity of the target image in vertical direction is determined based on the first edge figure, obtains second point Cut quantity；

Based on first dividing number and second dividing number, determine whether the target image is picture mosaic figure Picture.

Optionally, described that the divisible number of the target image in the horizontal direction is determined based on the first edge figure The step of measuring, obtaining the first dividing number, including：

The quantity for calculating object pixel row in the first edge figure, obtains the first quantity；The object pixel row indicates Including the number of object pixel be more than the pixel column of the first predetermined number threshold value, it is pre- that the object pixel indicates that gray value is equal to If the pixel of gray value；

If first quantity is more than the first predetermined threshold value, or, being less than the second predetermined threshold value, it is determined that first segmentation Quantity is 0；

If first quantity is no more than first predetermined threshold value and is not less than second predetermined threshold value, calculate every The distance between a object pixel row obtains multiple first distances；

By corresponding value maximum and highest first distance of the frequency of occurrences is as first object distance, if the first object The frequency of occurrences of distance is less than predeterminated frequency threshold value, it is determined that first dividing number is 0；

If the frequency of occurrences of the first object distance is not less than predeterminated frequency threshold value, in the first object apart from right When the distance value answered is less than pre-determined distance threshold value, determine that first dividing number is 0, in the first object apart from corresponding When distance value is not less than pre-determined distance threshold value, the frequency of occurrences of the first object distance is determined as the first segmentation number Amount.

Optionally, described based on determining the target image in vertical direction divisible described in the first edge figure Quantity, the step of obtaining the second dividing number, including：

The quantity for calculating the object pixel row for including in the second edge figure, obtains the second quantity；The object pixel Row indicate that the number for the object pixel for including is more than the pixel column of the second predetermined number threshold value, and the object pixel indicates gray value Equal to the pixel of default gray value；

If second quantity is more than first predetermined threshold value, or, being less than second predetermined threshold value, it is determined that described Second dividing number is 0；

If second quantity is no more than first predetermined threshold value and is not less than second predetermined threshold value, calculate every The distance between a object pixel row, obtain multiple second distances；

Using corresponding value maximum and the highest second distance of the frequency of occurrences is as the second target range, if second target The frequency of occurrences of distance is less than predeterminated frequency threshold value, it is determined that second dividing number is 0；

If the frequency of occurrences of second target range be not less than the predeterminated frequency threshold value, second target away from From corresponding distance value be less than the pre-determined distance threshold value when, determine second dividing number be 0, second target away from When being not less than the pre-determined distance threshold value from corresponding distance value, the frequency of occurrences of second target range is determined as described Second dividing number.

Optionally, if the target image is picture mosaic image, the target image is split, described in acquisition The step of at least two target subgraphs that target image includes, including：

If first dividing number and second dividing number are not zero, based on the object pixel row with And the object pixel row, the target image is split, multiple target subgraphs are obtained；

If first dividing number is not zero, second dividing number is zero, then is based on the object pixel row pair The target image is split, and obtains multiple second subgraphs；Target subgraph is determined based on the multiple second subgraph；

If second dividing number is not zero, first dividing number is zero, then is based on object pixel row pair The target image is split, and obtains multiple third subgraphs；Target subgraph is determined based on the multiple third subgraph.

Optionally, described the step of target subgraph is determined based on the multiple second subgraph, including：

For each second subgraph, determine that second subgraph in the edge graph of vertical direction, obtains third edge Figure；

Based on the third edge graph, judge that second subgraph whether there is divisible position in vertical direction；

If there are divisible position, the divisible position based on second subgraph in vertical direction, to described Second subgraph is split, and the subgraph obtained after segmentation is determined as target subgraph；

If divisible position is not present, second subgraph is determined as target subgraph.

Optionally, described the step of target subgraph is determined based on the multiple third subgraph, including：

For each third subgraph, determines the edge graph of the third subgraph in the horizontal direction, obtain the 4th side Edge figure；

Based on the 4th edge graph, judge that the third subgraph whether there is divisible position in the horizontal direction；

If there are divisible position, the divisible position based on the third subgraph in the horizontal direction, to described Third subgraph is split, and the subgraph obtained after segmentation is determined as target subgraph；

If divisible position is not present, the third subgraph is determined as target subgraph.

Second aspect according to the present invention provides a kind of image processing apparatus, is applied to terminal, which can wrap It includes：

Judgment module, for judging whether target image is picture mosaic image；

Divide module, if being picture mosaic image for the target image, the target image is split, obtains institute State at least two target subgraphs that target image includes；

Sort module, for classifying respectively to every target subgraph, and based on the classification of every target subgraph As a result the classification of the target image is determined.

Optionally, the judgment module, including：

Transform subblock, for the target image to be converted to gray level image, obtaining can processing target image；

First determination sub-module, for be based on preset horizontal edge operator, determine described in can processing target image in water Square to first edge figure, and, be based on preset vertical edge operator, determine described in can processing target image in Vertical Square To second edge figure；

Second determination sub-module, for based on the first edge figure determine the target image in the horizontal direction can Dividing number obtains the first dividing number, based on the first edge figure determine the target image in vertical direction can Dividing number obtains the second dividing number；

Third determination sub-module, described in based on first dividing number and second dividing number, determining Whether target image is picture mosaic image.

Optionally, second determination sub-module, is used for：

If first quantity is no more than second predetermined threshold value and is not less than first predetermined threshold value, calculate every The distance between a object pixel row obtains multiple first distances；

Optionally, second determination sub-module, is used for, including：

Optionally, the segmentation module, including：

First segmentation submodule, if being not zero for first dividing number and second dividing number, It is arranged based on the object pixel row and the object pixel, the target image is split, multiple target subgraphs are obtained Picture；

Second segmentation submodule, if being not zero for first dividing number, second dividing number is zero, then base The target image is split in the object pixel row, obtains multiple second subgraphs；4th determination sub-module, is used for Target subgraph is determined based on the multiple second subgraph；

Third divides submodule, if being not zero for second dividing number, first dividing number is zero, then base It is arranged in the object pixel and the target image is split, obtain multiple third subgraphs；5th determination sub-module, is used for Target subgraph is determined based on the multiple third subgraph.

Optionally, the 4th determination sub-module, is used for：

Optionally, the 5th determination sub-module, is used for：

For first technology, the present invention has following advantage：

Terminal can first judge whether target image is picture mosaic image, then, can be when target image is picture mosaic image, to mesh Logo image is split, and at least two target subgraphs that target image includes is obtained, finally, respectively to every target subgraph Classify, and determine the classification of target image based on the classification results of every target subgraph, in this way, by by target image It splits, reduces the content complexity of target image, the standard of judgement can be improved to avoid due to the complicated caused erroneous judgement of picture material True rate.

Above description is only the general introduction of technical solution of the present invention, in order to better understand the technical means of the present invention, And can be implemented in accordance with the contents of the specification, and in order to allow above and other objects of the present invention, feature and advantage can It is clearer and more comprehensible, below the special specific implementation mode for lifting the present invention.

Description of the drawings

By reading the detailed description of hereafter preferred embodiment, various other advantages and benefit are common for this field Technical staff will become clear.Attached drawing only for the purpose of illustrating preferred embodiments, and is not considered as to the present invention Limitation.And throughout the drawings, the same reference numbers will be used to refer to the same parts.In the accompanying drawings：

Fig. 1 is a kind of step flow chart for image processing method that the embodiment of the present invention one provides；

Fig. 2-1 is a kind of step flow chart of image processing method provided by Embodiment 2 of the present invention；

Fig. 2-2 is first edge schematic diagram provided by Embodiment 2 of the present invention；

Fig. 2-3 is second edge schematic diagram provided by Embodiment 2 of the present invention；

Fig. 2-4 is a kind of fractionation schematic diagram provided by Embodiment 2 of the present invention；

Fig. 2-5 is another fractionation schematic diagram provided by Embodiment 2 of the present invention；

Fig. 3 is a kind of block diagram for image processing apparatus that the embodiment of the present invention three provides；

Fig. 4 is a kind of block diagram for image processing apparatus that the embodiment of the present invention four provides.

Specific implementation mode

Exemplary embodiment of the present invention is more fully described below with reference to accompanying drawings.Although showing the present invention in attached drawing Exemplary embodiment, it being understood, however, that may be realized in various forms the present invention without should be by embodiments set forth here It is limited.It is to be able to be best understood from the present invention on the contrary, providing these embodiments, and can be by the scope of the present invention Completely it is communicated to those skilled in the art.

Embodiment one

Fig. 1 is a kind of step flow chart for image processing method that the embodiment of the present invention one provides, and is applied to terminal, such as schemes Shown in 1, this method may include：

Step 101 judges whether target image is picture mosaic image.

In the embodiment of the present invention, since the content of picture mosaic image is complex, the False Rate of classification is higher, and to be sorted Target image is likely to be picture mosaic image, and therefore, terminal can first judge target image before classifying to target image Whether it is picture mosaic image.Specifically, terminal can first obtain the edge graph of target image, wherein the edge graph can embody mesh The characteristics of image of logo image, the edge graph for being then based on target image determine divisible quantity, it is true to be finally based on divisible quantity Whether the image that sets the goal is picture mosaic image.

If step 102, the target image are picture mosaic image, the target image is split, the mesh is obtained At least two target subgraphs that logo image includes.

In the embodiment of the present invention, if target image is picture mosaic image, illustrate target image by the independent son of multiple contents Image forms, and in order to improve the judging nicety rate to target image classification, target image can be divided into sub one by one by terminal Image obtains multiple target subgraphs, and then reduces the content complexity of target image, improves subsequent classification accuracy.Into One step, if target image is not picture mosaic image, terminal directly can utilize preset disaggregated model to the target figure As classifying.

Step 103 respectively classifies to every target subgraph, and the classification results based on every target subgraph are true The classification of the fixed target image.

In the embodiment of the present invention, since every target subgraph is all a part for target image, that is, every target subgraph A part of content as all representing target image, therefore, terminal can first judge the classification of each target subgraph, be finally based on The classification of every target subgraph determines the classification of target image, in this manner it is ensured that when determining target image classification, energy Enough fully take into account all the elements of target image.

In conclusion the image processing method that the embodiment of the present invention one provides, can first judge whether target image is picture mosaic Image can then be split target image, it includes at least to obtain target image when target image is picture mosaic image Two target subgraphs finally respectively classify to every target subgraph, and based on the classification knot of every target subgraph Fruit determines the classification of target image, in this way, by splitting target image, reduces the content complexity of target image, can keep away Exempt from, due to the complicated caused erroneous judgement of picture material, to improve the accuracy rate of judgement.

Embodiment two

Fig. 2-1 is a kind of step flow chart of image processing method provided by Embodiment 2 of the present invention, is applied to terminal, such as Shown in Fig. 2-1, this method may include：

The target image is converted to gray level image by step 201, and obtaining can processing target image.

It, can be by carrying out gray processing processing to target image, to obtain the corresponding gray-scale map of target image in this step Picture, wherein gray processing processing is the process that coloured image is converted to gray level image, carries out the gray scale obtained after gray processing processing Image can reflect distribution and the feature of the entirety and local coloration and brightness degree of entire image.The embodiment of the present invention In, by the way that colored target image is converted to gray level image, calculating quantitative change when can make subsequently for target image processing It must lack, and then reduce the consuming to terminal system resource.Specifically, terminal can be based on the red color channel value of each pixel R, green channel value G and blue channel value B calculates the corresponding gray value Gray of each pixel by following formula 1, then will The gray scale of each pixel is set as the corresponding gray value of each pixel, obtains gray level image.

Gray=R*0.299+G*0.587+B*0.114；(formula 1)

It should be noted that in another alternative embodiment of the present invention, the maximum R that is also based in target image Value, maximum G values, maximum B values, minimum R values, minimum G values and minimum B values calculate a normal brightness value using following formula 2 Then L sets the brightness value of each pixel to the normal brightness value, obtains gray level image.

L=(max (R, G, B)+min (R, G, B))/2；(formula 2)

Step 202 is based on preset horizontal edge operator, can processing target image in the horizontal direction first described in determination Edge graph, and, be based on preset vertical edge operator, determine described in can processing target image vertical direction second edge Figure.

In this step, the horizontal edge operator and vertical edge operator can be Sobel Sobel operators, specifically, The horizontal edge operator can be the minor matrix of a 3*3：

-1	0	+1
			-2	0	+2
-1	0	+1

Further, which can be the minor matrix of a 3*3：

+1	+2	+1
			0	0	0
-1	-2	-1

Further, terminal can using can be in processing target image each pixel as can processing target image it is corresponding Each element in matrix, the gray value of each pixel are the value of each element, then, for can processing target image correspond to Matrix in each element, calculate the element and around it 8 pixels with it is right in the minor matrix of horizontal edge operator representation The product of position element is answered, then by the addition of obtained product, the new gray value of the corresponding pixel of the element is obtained, with this Analogize, can be in processing target image after the new gray value of each pixel calculating, it can be in processing target image The gray value of each pixel is set as the corresponding new gray value of each pixel, finally, to after adjustment can processing target image Carry out binary conversion treatment, obtain can processing target image first edge figure in the horizontal direction, it is exemplary, Fig. 2-2 be the present invention The first edge schematic diagram that embodiment two provides.

Correspondingly, terminal can using can be in processing target image each pixel as can the corresponding square of processing target image Battle array in each element, the gray value of each pixel is the value of each element, then, for can processing target image it is corresponding Each element in matrix calculates the element and 8 pixels is corresponding with the minor matrix that vertical edge operator indicates around it The product of position element obtains the new gray value of the corresponding pixel of the element, with such then by the addition of obtained product It pushes away, can be in processing target image after the new gray value of each pixel calculating, it can be every in processing target image The gray value of a pixel is set as the corresponding new gray value of each pixel, finally, to after adjustment can processing target image into Row binary conversion treatment, obtain can processing target image second edge figure in the horizontal direction, exemplary, Fig. 2-3 is that the present invention is real The second edge schematic diagram of the offer of example two is provided.

Specifically, the binary conversion treatment is exactly that the gray value of each pixel in the gray level image is set as 0 or 255, make Whole image shows apparent black and white effect, can should specifically, be less than the pixel of default gray threshold for gray value The gray value of pixel is set as 0, and the pixel of default gray threshold is not less than for gray value, can set the gray value of the pixel 255 are set to, which can be selected according to actual conditions, exemplary, which can be 60, The default gray threshold may be 80, and the embodiment of the present invention is not construed as limiting this.

The binary image obtained after binary conversion treatment remains able to reflection image entirety and local feature.Further Ground, since the gray value of pixel in the image after binary conversion treatment only has 0 or 255, the rank of grey scale pixel value is few, therefore number According to processing and decrement it is small, and then handled again after target image is converted to binary image, processing can be made to grasp Make simpler.

Step 203 determines the divisible quantity of the target image in the horizontal direction based on the first edge figure, obtains To the first dividing number, the divisible quantity of the target image in vertical direction is determined based on the first edge figure, is obtained To the second dividing number.

Specifically, terminal can realize determining first edge figure in level by following sub-steps (1)~sub-step (5) Divisible quantity on direction, obtains the first dividing number：

Sub-step (1)：The quantity for calculating object pixel row in the first edge figure, obtains the first quantity.

In this step, object pixel row can indicate that the number for the object pixel for including is more than the first predetermined number threshold value Pixel column, the object pixel indicate that gray value is equal to the pixel of default gray value, which can be 255.Specifically, Terminal can count the number of pixels that gray value included in each pixel column in first edge figure is equal to default gray value, so That included gray value is more than the pixel column of the first predetermined number threshold value equal to the number of pixels of default gray value is calculated afterwards Number, obtains the first quantity.Wherein, which can be the pixel included by pixel column in target image What number determined, when the number of pixels included by the pixel column in target image is more, default of larger first can be set The first smaller predetermined number threshold can be arranged when the number of pixels included by the pixel column in target image is less in number threshold values Value.

Sub-step (2)：If first quantity is more than the first predetermined threshold value, or, being less than the second predetermined threshold value, it is determined that institute It is 0 to state the first dividing number.

In this step, the first predetermined threshold value is more than the second predetermined threshold value, further, in actual application scenarios, picture mosaic There can be the more moderate part of edge strength in the content of image, if gray value included in image is equal to default gray scale When the number of pixels of value is excessive, it may be considered that the edge strength in the content of the image is too strong, if included by image When gray value is very few equal to the number of pixels of default gray value, it may be considered that the edge strength in the content of the image is excessively weak, At this time it is considered that the image is not picture mosaic image, that is, the image can not be split.

Correspondingly, in the embodiment of the present invention, if the first quantity is more than the first predetermined threshold value, it may be considered that first edge figure Edge it is too strong, that is, the edge of target image horizontal direction is too strong；If the first quantity is less than the second predetermined threshold value, can recognize Edge for first edge figure is excessively weak, that is, the edge of target image horizontal direction is excessively weak, and target image does not have in the horizontal direction The position that can divide, hence, it can be determined that the first dividing number of target image in the horizontal direction is 0.

Sub-step (3)：If first quantity is no more than first predetermined threshold value and is not less than the described second default threshold Value, then calculate the distance between each object pixel row, obtain multiple first distances.

It is exemplary, it is assumed that object pixel row includes a rows, b rows, c rows, e rows and f rows, then terminal can With calculate a rows at a distance from b rows, a rows at a distance from c rows, a rows are at a distance from e rows, a rows and f rows Distance, b rows at a distance from c rows, b rows at a distance from e rows, b rows are at a distance from f rows, c rows and e rows Distance, c rows are at a distance from f rows and e rows are at a distance from f rows, obtain 10 the first distances.

Sub-step (4)：By corresponding value maximum and highest first distance of the frequency of occurrences is as first object distance, if institute The frequency of occurrences for stating first object distance is less than predeterminated frequency threshold value, it is determined that first dividing number is 0.

In this step, which can be set according to actual demand, and the embodiment of the present invention does not limit this It is fixed.It is exemplary, it is assumed that predeterminated frequency threshold value is 3, the frequency of occurrences 2 of first object distance, due to the appearance of first object distance Frequency is less than predeterminated frequency threshold value, then terminal can determine that the first dividing number is 0.Assuming that predeterminated frequency threshold value is 3, first The frequency of occurrences of target range is 3, since the frequency of occurrences of first object distance is not less than predeterminated frequency threshold value, then terminal can To determine the first dividing number not for 0.

Sub-step (5)：If the frequency of occurrences of the first object distance is not less than predeterminated frequency threshold value, described first When the corresponding distance value of target range is less than pre-determined distance threshold value, determine that first dividing number is 0, in the first object When being not less than pre-determined distance threshold value apart from corresponding distance value, the frequency of occurrences of the first object distance is determined as described the One dividing number.

In this step, which can be set according to actual demand, and the embodiment of the present invention does not limit this It is fixed.It is exemplary, it is assumed that pre-determined distance threshold value be 3 centimetres of cm, first object apart from corresponding distance value be 5cm, first object away from From the frequency of occurrences be 3, due to first object apart from corresponding distance value be not less than pre-determined distance threshold value, then terminal can be true Fixed first dividing number is 3.

Specifically, terminal can realize that determining second edge figure is hanging down by following sub-steps (6)~sub-step (10) The upward divisible quantity of histogram, obtains the second dividing number：

Sub-step (6)：The quantity for calculating the object pixel row for including in the second edge figure, obtains the second quantity.

In this step, object pixel row can indicate that the number for the object pixel for including is more than the second predetermined number threshold value Pixel column, the object pixel indicate that gray value is equal to the pixel of default gray value, which can be 255.Specifically, Terminal can count the number of pixels that gray value included in each pixel column in second edge figure is equal to default gray value, so That included gray value is more than the pixel column of the second predetermined number threshold value equal to the number of pixels of default gray value is calculated afterwards Number, obtains the second quantity.Wherein, which can be the pixel included by pixel column in target image What number determined, when the number of pixels included by the pixel column in target image is more, default of larger second can be set The second smaller predetermined number threshold can be arranged when the number of pixels included by the pixel column in target image is less in number threshold values Value.

Sub-step (7)：If second quantity is more than first predetermined threshold value, or, it is less than second predetermined threshold value, Then determine that second dividing number is 0.

There can be the more moderate part of edge strength in actual application scenarios, in the content of picture mosaic image, if figure When included gray value is excessive equal to the number of pixels of default gray value as in, it may be considered that the side in the content of the image Edge intensity is too strong, if gray value included in image is very few equal to the number of pixels of default gray value, it may be considered that Edge strength in the content of the image is excessively weak, at this time it is considered that the image is not picture mosaic image, that is, the image can not carry out Segmentation.

Correspondingly, in the embodiment of the present invention, if the second quantity is more than the first predetermined threshold value, it may be considered that second edge figure Edge it is too strong, that is, the edge of target image vertical direction is too strong；If the second quantity is less than the second predetermined threshold value, can recognize Edge for second edge figure is excessively weak, that is, the edge of target image vertical direction is excessively weak, and target image does not have in vertical direction The position that can divide, hence, it can be determined that the second dividing number of target image in vertical direction is 0.

Sub-step (8)：If second quantity is no more than first predetermined threshold value and is not less than the described second default threshold Value then calculates each object pixel row and adjacent the distance between object pixel row, obtains multiple second distances.

It is exemplary, it is assumed that object pixel row include h row, the i-th row, the i-th row, kth row and the 1st row, then terminal can With calculate h row with i-th row at a distance from, h row with jth row at a distance from, h row with kth row at a distance from, h arrange and the 1st arrange Distance, i-th row with jth row at a distance from, i-th row with kth row at a distance from, i-th row with the 1st row at a distance from, jth arrange and kth arrange Distance, jth row are at a distance from the 1st row and kth row are at a distance from the 1st row, obtain 10 second distances.

Sub-step (9)：Using corresponding value maximum and the highest second distance of the frequency of occurrences is as the second target range, if institute The frequency of occurrences for stating the second target range is less than predeterminated frequency threshold value, it is determined that second dividing number is 0.

In this step, which can be set according to actual demand, and the embodiment of the present invention does not limit this It is fixed.It is exemplary, it is assumed that predeterminated frequency threshold value is 3, the frequency of occurrences 2 of the second target range, due to the appearance of the second target range Frequency is less than predeterminated frequency threshold value, then terminal can determine that the second dividing number is 0.Assuming that predeterminated frequency threshold value is 3, second The frequency of occurrences of target range is 4, since the frequency of occurrences of the second target range is not less than predeterminated frequency threshold value, then terminal can To determine the second dividing number not for 0.

Sub-step (10)：If the frequency of occurrences of second target range is not less than the predeterminated frequency threshold value, in institute When stating the corresponding distance value of the second target range less than the pre-determined distance threshold value, determine that second dividing number is 0, in institute When stating the corresponding distance value of the second target range not less than the pre-determined distance threshold value, by the appearance frequency of second target range Rate is determined as second dividing number.

In this step, which can be set according to actual demand, and the embodiment of the present invention does not limit this It is fixed.It is exemplary, it is assumed that pre-determined distance threshold value be 3 centimetres of cm, the corresponding distance value of the second target range be 8cm, the second target away from From the frequency of occurrences be 4, due to the corresponding distance value of the second target range be not less than pre-determined distance threshold value, then terminal can be true Fixed second dividing number is 4.

Step 204 is based on first dividing number and second dividing number, whether determines the target image For picture mosaic image.

If specifically, first dividing number and second dividing number are zero, it is determined that the target figure As not being picture mosaic image.If first dividing number and second dividing number are not zero there are at least one, for example, First dividing number is not 0, alternatively, the second dividing number is not 0, alternatively, the first dividing number and the second dividing number are equal It is not 0, then can determines that target image is picture mosaic image.

If step 205, the target image are picture mosaic image, the target image is split, the mesh is obtained At least two target subgraphs that logo image includes.

Specifically, terminal can divide target image to realize by following sub-steps (11)~sub-step (13) It cuts, obtains at least two target subgraphs that target image includes：

Sub-step (11)：If first dividing number and second dividing number are not zero, based on described Object pixel row and object pixel row, are split the target image, obtain multiple target subgraphs.

In this step, terminal can carry out horizontal direction first along object pixel row corresponding row in the target image Segmentation arranges corresponding row in the target image then along object pixel, the segmentation of vertical direction is carried out, finally, after segmentation Obtained each subgraph, is determined as target subgraph.Certainly, in practical application, when being split, can also be first along Object pixel arranges corresponding row in the target image, the segmentation of vertical direction is carried out, then along object pixel row in target figure The corresponding row as in, carries out the segmentation of horizontal direction, the embodiment of the present invention is not construed as limiting this.Exemplary, Fig. 2-4 is the present invention A kind of fractionation schematic diagram that embodiment two provides, as can be seen that " artwork " is represented from " split result " shown in Fig. 2-4 Target image be split as 9 target subgraphs.Sub-step (12)：If first dividing number is not zero, described second Dividing number is zero, then is split to the target image based on the object pixel row, obtains multiple second subgraphs；Base Target subgraph is determined in the multiple second subgraph.

In this step, when the first dividing number is not 0, it is believed that target image is divisible in the horizontal direction, this When, terminal can carry out the segmentation of horizontal direction, by target image first along object pixel row corresponding row in the target image Multiple second subgraphs are divided into, due to that in practical application, there may be irregular picture mosaic, that is, there is only parts second Subgraph is divisible in vertical direction, therefore, in order to ensure picture mosaic segmentation accuracy, terminal can by following procedure, Each second subgraph is handled, to determine final target subgraph：

Firstly, for each second subgraph, determine that second subgraph in the edge graph of vertical direction, obtains third Edge graph.

Specifically, determine that the realization process of third edge graph can refer to above-mentioned steps 202, the embodiment of the present invention is herein not It repeats.

It is then possible to be based on the third edge graph, judge second subgraph in vertical direction with the presence or absence of can Split position.

Specifically, the realization process of this step can refer to the process that above-mentioned sub-step (6)~sub-step (10) is shown, this This will not be repeated here for inventive embodiments.

Finally, if there are divisible position, the divisible position based on second subgraph in vertical direction is right Second subgraph is split, and the subgraph obtained after segmentation is determined as target subgraph, if there is no divisible Second subgraph is then determined as target subgraph by position.

It is exemplary, Fig. 2-5 be it is provided by Embodiment 2 of the present invention it is another split schematic diagram, " tear open shown in Fig. 2-5 As can be seen that the target image represented by " artwork " is split as 6 target subgraphs in point result ".

Sub-step (13)：If second dividing number is not zero, first dividing number is zero, then is based on the mesh Mark pixel column is split the target image, obtains multiple third subgraphs；It is determined based on the multiple third subgraph Target subgraph.

In this step, when the second dividing number is not 0, it is believed that target image is divisible in vertical direction, this When, terminal first can arrange corresponding row in the target image along object pixel, the segmentation of vertical direction be carried out, by target image Multiple third subgraphs are divided into, due to that in practical application, there may be irregular picture mosaic, that is, there is only part thirds Subgraph is divisible in the horizontal direction, therefore, in order to ensure picture mosaic segmentation accuracy, terminal can by following procedure, Each third subgraph is handled, to determine final target subgraph：

Firstly, for each third subgraph, the edge graph of the third subgraph in the horizontal direction is determined, obtain Four edge graphs.

Specifically, determining that the realization process of the 4th edge graph can refer to above-mentioned steps 202, the embodiment of the present invention is herein not It repeats.

Then, it is based on the 4th edge graph, it is divisible to judge that the third subgraph whether there is in the horizontal direction Position.

Specifically, the realization process of this step can refer to the process that above-mentioned sub-step (1)~sub-step (5) is shown, this This will not be repeated here for inventive embodiments.

Finally, if there are divisible position, the divisible position based on the third subgraph in the horizontal direction is right The third subgraph is split, and the subgraph obtained after segmentation is determined as target subgraph, if there is no divisible The third subgraph is then determined as target subgraph by position.

It should be noted that in order to further improve the accuracy rate that picture mosaic image is split, in the another optional of the present invention In embodiment, terminal can also continue to each target subgraph execute judge the target subgraph whether be picture mosaic image behaviour Make, if the target subgraph is picture mosaic image, target subgraph is continued to divide, until dividing obtained every height When image is not picture mosaic image, then based on each classification divided obtained subgraph and determine target image.

Step 206 respectively classifies to every target subgraph, and the classification results based on every target subgraph are true The classification of the fixed target image.

Exemplary, for judging whether target image is pornographic image, terminal can collect a large amount of normogram in advance Then picture and pornographic image manually mark out the classification of each sample image, that is, marking out the sample image is as sample Normal picture or pornographic image, then, by these sample images and and each sample image mark classification, be input to Model is carried out in the deep learning frame that caffe increases income to instruct again, other deep learnings of increasing income can also be used certainly, in practical application Frame is trained, and the embodiment of the present invention is not construed as limiting this.

Specifically in training, feedforward BP neural network algorithm may be used, i.e., it is preceding to be exported in layer to when transmitting, if The result that output layer obtains has difference then to carry out back transfer with desired value, updates it with gradient descent method according to its error Weight and threshold values, repeated several times, until error function reaches global minimum, end training obtains pornographic image classification mould Type.

In classification, every target subgraph can be input in the pornographic image disaggregated model, according to same mould Type structure and trained parameter are successively handled every target subgraph, for example, convolution, pond, etc., until The confidence level probability that every target subgraph corresponds to two classifications (normal picture classification, pornographic image) is finally obtained, it finally, will The corresponding classification of maximum confidence probability is determined as the classification of target image.

Certainly, in another alternative embodiment of the present invention, two can also be corresponded to acquiring every target subgraph After the confidence level probability of classification, maximum confidence probability in the corresponding two confidence level probability of each target subgraph is corresponded to Classification, be determined as the classification of the target subgraph, finally, when the classification of all target subgraphs is normal picture, really The image that sets the goal is normal picture, when the classification of at least one target subgraph is pornographic image, determines that target image is color Feelings image.Further, in order to improve the accuracy rate of judgement, terminal can also to target subgraph that classification is pornographic image into Rower is noted, in order to manually be checked.

In conclusion image processing method provided by Embodiment 2 of the present invention, first can determine target image in the horizontal direction First edge figure and second edge figure in vertical direction, be then based on first edge figure and second edge figure determine mesh The first dividing number and the second dividing number of logo image, are then based on the first dividing number and the second dividing number judges Whether target image is picture mosaic image, then can be split to target image when target image is picture mosaic image, obtain mesh At least two target subgraphs that logo image includes finally respectively classify to every target subgraph, and are based on every mesh The classification results of mark subgraph determine that the classification of target image reduces the interior of target image in this way, by splitting target image Hold complexity, the accuracy rate of judgement can be improved to avoid due to the complicated caused erroneous judgement of picture material.

Embodiment three

Fig. 3 is a kind of block diagram for image processing apparatus that the embodiment of the present invention three provides, as shown in figure 3, the device 30 can To include：

Judgment module 301, for judging whether target image is picture mosaic image.

Divide module 302, if being picture mosaic image for the target image, the target image is split, is obtained Take at least two target subgraphs that the target image includes.

Sort module 303, for classifying respectively to every target subgraph, and based on point of every target subgraph Class result determines the classification of the target image.

In conclusion the image processing apparatus that the embodiment of the present invention three provides, judgment module can first judge that target image is No is picture mosaic image, and then, segmentation module can be split target image, obtain mesh when target image is picture mosaic image At least two target subgraphs that logo image includes, finally, sort module can respectively classify to every target subgraph, and The classification of target image is determined based on the classification results of every target subgraph, in this way, by splitting target image, reduces mesh The content complexity of logo image can improve the accuracy rate of judgement to avoid due to the complicated caused erroneous judgement of picture material.

Example IV

Fig. 4 is a kind of block diagram for image processing apparatus that the embodiment of the present invention four provides, as shown in figure 4, the device 40 can To include：

Judgment module 401, for judging whether target image is picture mosaic image.

Divide module 402, if being picture mosaic image for the target image, the target image is split, is obtained Take at least two target subgraphs that the target image includes.

Sort module 403, for classifying respectively to every target subgraph, and based on point of every target subgraph Class result determines the classification of the target image.

Optionally, above-mentioned judgment module 401, including：

Transform subblock 4011, for the target image to be converted to gray level image, obtaining can processing target image.

First determination sub-module 4012 can processing target image described in determination for being based on preset horizontal edge operator First edge figure in the horizontal direction, and, be based on preset vertical edge operator, determine described in can processing target image hang down Histogram to second edge figure.

Second determination sub-module 4013, for determining the target image in the horizontal direction based on the first edge figure Divisible quantity, obtain the first dividing number, the target image determined in vertical direction based on the first edge figure Divisible quantity, obtain the second dividing number.

Third determination sub-module 4014 is determined for being based on first dividing number and second dividing number Whether the target image is picture mosaic image.

Optionally, above-mentioned second determination sub-module 4013, is used for：

The quantity for calculating object pixel row in the first edge figure, obtains the first quantity；The object pixel row indicates Including the number of object pixel be more than the pixel column of the first predetermined number threshold value, it is pre- that the object pixel indicates that gray value is equal to If the pixel of gray value.

If first quantity is more than the first predetermined threshold value, or, being less than the second predetermined threshold value, it is determined that first segmentation Quantity is 0.

If first quantity is no more than first predetermined threshold value and is not less than second predetermined threshold value, calculate every The distance between a object pixel row obtains multiple first distances.

By corresponding value maximum and highest first distance of the frequency of occurrences is as first object distance, if the first object The frequency of occurrences of distance is less than predeterminated frequency threshold value, it is determined that first dividing number is 0.

Optionally, above-mentioned second determination sub-module 4013, is used for, including：

The quantity for calculating the object pixel row for including in the second edge figure, obtains the second quantity；The object pixel Row indicate that the number for the object pixel for including is more than the pixel column of the second predetermined number threshold value, and the object pixel indicates gray value Equal to the pixel of default gray value.

If second quantity is more than first predetermined threshold value, or, being less than second predetermined threshold value, it is determined that described Second dividing number is 0.

If second quantity is no more than first predetermined threshold value and is not less than second predetermined threshold value, calculate every The distance between a object pixel row, obtain multiple second distances.

Using corresponding value maximum and the highest second distance of the frequency of occurrences is as the second target range, if second target The frequency of occurrences of distance is less than predeterminated frequency threshold value, it is determined that second dividing number is 0.

Optionally, above-mentioned third determination sub-module 4014, is used for：

If first dividing number and second dividing number are zero, it is determined that the target image is not to spell Figure image.

If first dividing number and second dividing number are not zero there are at least one, it is determined that the mesh Logo image is picture mosaic image.

Optionally, above-mentioned segmentation module 402, including：

First segmentation submodule, if being not zero for first dividing number and second dividing number, It is arranged based on the object pixel row and the object pixel, the target image is split, multiple target subgraphs are obtained Picture.

Second segmentation submodule, if being not zero for first dividing number, second dividing number is zero, then base The target image is split in the object pixel row, obtains multiple second subgraphs；4th determination sub-module, is used for Target subgraph is determined based on the multiple second subgraph.

Optionally, above-mentioned 4th determination sub-module, is used for：

For each second subgraph, determine that second subgraph in the edge graph of vertical direction, obtains third edge Figure.

Based on the third edge graph, judge that second subgraph whether there is divisible position in vertical direction.

If there are divisible position, the divisible position based on second subgraph in vertical direction, to described Second subgraph is split, and the subgraph obtained after segmentation is determined as target subgraph.

Optionally, above-mentioned 5th determination sub-module, is used for：

For each third subgraph, determines the edge graph of the third subgraph in the horizontal direction, obtain the 4th side Edge figure.

Based on the 4th edge graph, judge that the third subgraph whether there is divisible position in the horizontal direction.

If there are divisible position, the divisible position based on the third subgraph in the horizontal direction, to described Third subgraph is split, and the subgraph obtained after segmentation is determined as target subgraph.

In conclusion the image processing apparatus that the embodiment of the present invention four provides, the first determination sub-module can first determine target Image first edge figure in the horizontal direction and the second edge figure in vertical direction, then the second determination sub-module can be based on First edge figure and second edge figure determine the first dividing number and the second dividing number of target image, and then third is true Stator modules can be based on the first dividing number and the second dividing number judges whether target image is picture mosaic image, then divide Module can be split target image when target image is picture mosaic image, obtain at least two mesh that target image includes Subgraph is marked, finally, sort module can respectively classify to every target subgraph, and based on point of every target subgraph Class result determines the classification of target image, in this way, by splitting target image, reduces the content complexity of target image, can To avoid due to the complicated caused erroneous judgement of picture material, the accuracy rate of judgement is improved.

For above-mentioned apparatus embodiment, since it is basically similar to the method embodiment, so description is fairly simple, The relevent part can refer to the partial explaination of embodiments of method.

Each embodiment in this specification is described in a progressive manner, the highlights of each of the examples are with The difference of other embodiment, the same or similar parts between the embodiments can be referred to each other.

It would have readily occurred to a person skilled in the art that be：The arbitrary combination application of above-mentioned each embodiment is all feasible, therefore Arbitrary combination between above-mentioned each embodiment is all embodiment of the present invention, but this specification exists as space is limited, This is not just detailed one by one.

Provided herein image processing method not with the intrinsic phase of any certain computer, virtual system or miscellaneous equipment It closes.Various general-purpose systems can also be used together with teaching based on this.As described above, construction has present invention side Structure required by the system of case is obvious.In addition, the present invention is not also directed to any certain programmed language.It should be bright In vain, various programming languages can be utilized to realize the content of invention described herein, and is retouched above to what language-specific was done State is to disclose the preferred forms of the present invention.

In the instructions provided here, numerous specific details are set forth.It is to be appreciated, however, that the implementation of the present invention Example can be put into practice without these specific details.In some instances, well known method, structure is not been shown in detail And technology, so as not to obscure the understanding of this description.

Similarly, it should be understood that in order to simplify the present invention and help to understand one or more of each inventive aspect, Above in the description of exemplary embodiment of the present invention, each feature of the invention is grouped together into single implementation sometimes In example, figure or descriptions thereof.However, the method for the disclosure should be construed to reflect following intention：It is i.e. required to protect Shield the present invention claims the more features of feature than being expressly recited in each claim.More precisely, such as right As claim reflects, inventive aspect is all features less than single embodiment disclosed above.Therefore, it then follows tool Thus claims of body embodiment are expressly incorporated in the specific implementation mode, wherein each claim conduct itself The separate embodiments of the present invention.

Those skilled in the art, which are appreciated that, to carry out adaptively the module in the equipment in embodiment Change and they are arranged in the one or more equipment different from the embodiment.It can be the module or list in embodiment Member or component be combined into a module or unit or component, and can be divided into addition multiple submodule or subelement or Sub-component.Other than such feature and/or at least some of process or unit exclude each other, it may be used any Combination is disclosed to all features disclosed in this specification (including adjoint claim, abstract and attached drawing) and so to appoint Where all processes or unit of method or equipment are combined.Unless expressly stated otherwise, this specification (including adjoint power Profit requires, abstract and attached drawing) disclosed in each feature can be by providing the alternative features of identical, equivalent or similar purpose come generation It replaces.

In addition, it will be appreciated by those of skill in the art that although some embodiments described herein include other embodiments In included certain features rather than other feature, but the combination of the feature of different embodiments means in of the invention Within the scope of and form different embodiments.For example, in detail in the claims, embodiment claimed it is one of arbitrary It mode can use in any combination.

The all parts embodiment of the present invention can be with hardware realization, or to run on one or more processors Software module realize, or realized with combination thereof.It will be understood by those of skill in the art that can use in practice Microprocessor or digital signal processor (DSP) come realize in image processing method according to the ... of the embodiment of the present invention some or The some or all functions of person's whole component.The present invention is also implemented as one for executing method as described herein Divide either whole equipment or program of device (for example, computer program and computer program product).Such this hair of realization Bright program can may be stored on the computer-readable medium, or can be with the form of one or more signal.It is such Signal can be downloaded from internet website and be obtained, and either provided on carrier signal or provided in any other forms.

It should be noted that the present invention will be described rather than limits the invention for above-described embodiment, and ability Field technique personnel can design alternative embodiment without departing from the scope of the appended claims.In the claims, Any reference mark between bracket should not be configured to limitations on claims.Word "comprising" does not exclude the presence of not Element or step listed in the claims.Word "a" or "an" before element does not exclude the presence of multiple such Element.The present invention can be by means of including the hardware of several different elements and being come by means of properly programmed computer real It is existing.In the unit claims listing several devices, several in these devices can be by the same hardware branch To embody.The use of word first, second, and third does not indicate that any sequence.These words can be explained and be run after fame Claim.

Claims

1. a kind of image processing method, which is characterized in that it is applied to terminal, the method includes：

Judge whether target image is picture mosaic image；

If the target image is picture mosaic image, the target image is split, obtaining the target image includes At least two target subgraphs；

Classify respectively to every target subgraph, and the target figure is determined based on the classification results of every target subgraph The classification of picture.

2. according to the method described in claim 1, it is characterized in that, it is described judge target image whether be picture mosaic image step Suddenly, including：

Based on preset horizontal edge operator, determine described in can processing target image first edge figure in the horizontal direction, and, Based on preset vertical edge operator, determine described in can processing target image vertical direction second edge figure；

The divisible quantity of the target image in the horizontal direction is determined based on the first edge figure, obtains the first segmentation number Amount, determines the divisible quantity of the target image in vertical direction based on the second edge figure, obtains the second segmentation number Amount；

Based on first dividing number and second dividing number, determine whether the target image is picture mosaic image.

3. according to the method described in claim 2, it is characterized in that, described determine the target figure based on the first edge figure As divisible quantity in the horizontal direction, the step of obtaining the first dividing number, including：

The quantity for calculating object pixel row in the first edge figure, obtains the first quantity；The object pixel row indicates The number of object pixel be more than the pixel column of the first predetermined number threshold value, the object pixel indicates that gray value is equal to default ash The pixel of angle value；

If first quantity is more than the first predetermined threshold value, or, being less than the second predetermined threshold value, it is determined that first dividing number It is 0；

If first quantity is no more than first predetermined threshold value and is not less than second predetermined threshold value, each mesh is calculated The distance between pixel column is marked, multiple first distances are obtained；

By corresponding value maximum and highest first distance of the frequency of occurrences is as first object distance, if the first object distance The frequency of occurrences be less than predeterminated frequency threshold value, it is determined that first dividing number be 0；

If the frequency of occurrences of the first object distance is not less than predeterminated frequency threshold value, in the first object apart from corresponding When distance value is less than pre-determined distance threshold value, determine that first dividing number is 0, in the first object apart from corresponding distance When value is not less than pre-determined distance threshold value, the frequency of occurrences of the first object distance is determined as first dividing number.

4. according to the method described in claim 3, it is characterized in that, described based on the determining mesh described in the second edge figure The divisible quantity of logo image in vertical direction, the step of obtaining the second dividing number, including：

The quantity for calculating the object pixel row for including in the second edge figure, obtains the second quantity；The object pixel list Show that the number for the object pixel for including is more than the pixel column of the second predetermined number threshold value, the object pixel indicates that gray value is equal to The pixel of default gray value；

If second quantity is no more than first predetermined threshold value and is not less than second predetermined threshold value, each mesh is calculated The distance between pixel column is marked, multiple second distances are obtained；

Using corresponding value maximum and the highest second distance of the frequency of occurrences is as the second target range, if second target range The frequency of occurrences be less than predeterminated frequency threshold value, it is determined that second dividing number be 0；

If the frequency of occurrences of second target range is not less than the predeterminated frequency threshold value, in second target range pair When the distance value answered is less than the pre-determined distance threshold value, determine that second dividing number is 0, in second target range pair When the distance value answered is not less than the pre-determined distance threshold value, the frequency of occurrences of second target range is determined as described second Dividing number.

5. if according to the method described in claim 2, it is characterized in that, the target image is picture mosaic image, to institute The step of stating target image to be split, obtaining at least two target subgraphs that the target image includes, including：

If first dividing number and second dividing number are not zero, it is based on the object pixel row and institute Object pixel row are stated, the target image is split, multiple target subgraphs are obtained；

If first dividing number is not zero, second dividing number is zero, then is based on the object pixel row to described Target image is split, and obtains multiple second subgraphs；Target subgraph is determined based on the multiple second subgraph；

If second dividing number is not zero, first dividing number is zero, then is arranged to described based on the object pixel Target image is split, and obtains multiple third subgraphs；Target subgraph is determined based on the multiple third subgraph.

6. according to the method described in claim 5, it is characterized in that, described determine target based on the multiple second subgraph The step of image, including：

For each second subgraph, determine that second subgraph in the edge graph of vertical direction, obtains third edge graph；

7. according to the method described in claim 5, it is characterized in that, described determine target based on the multiple third subgraph The step of image, including：

For each third subgraph, determines the edge graph of the third subgraph in the horizontal direction, obtain the 4th edge graph；

If there are divisible position, the divisible position based on the third subgraph in the horizontal direction, to the third Subgraph is split, and the subgraph obtained after segmentation is determined as target subgraph；

8. a kind of image processing apparatus, which is characterized in that be applied to terminal, described device includes：

Judgment module, for judging whether target image is picture mosaic image；

Divide module, if being picture mosaic image for the target image, the target image is split, the mesh is obtained At least two target subgraphs that logo image includes；

Sort module, for classifying respectively to every target subgraph, and based on the classification results of every target subgraph Determine the classification of the target image.

9. device according to claim 8, which is characterized in that the judgment module, including：

First determination sub-module, for be based on preset horizontal edge operator, determine described in can processing target image in level side To first edge figure, and, be based on preset vertical edge operator, determine described in can processing target image in vertical direction Second edge figure；

Second determination sub-module, for determining the target image in the horizontal direction divisible based on the first edge figure Quantity obtains the first dividing number, and the target image in vertical direction divisible is determined based on the first edge figure Quantity obtains the second dividing number；

Third determination sub-module determines the target for being based on first dividing number and second dividing number Whether image is picture mosaic image.

10. device according to claim 9, which is characterized in that second determination sub-module is used for：

11. device according to claim 9, which is characterized in that second determination sub-module is used for：

12. device according to claim 9, which is characterized in that the segmentation module, including：

First segmentation submodule is based on if being not zero for first dividing number and second dividing number The object pixel row and object pixel row, are split the target image, obtain multiple target subgraphs；

Second segmentation submodule, if being not zero for first dividing number, second dividing number is zero, then is based on institute It states object pixel row to be split the target image, obtains multiple second subgraphs；4th determination sub-module, for being based on The multiple second subgraph determines target subgraph；

Third divides submodule, if being not zero for second dividing number, first dividing number is zero, then is based on institute It states object pixel and arranges and the target image is split, obtain multiple third subgraphs；5th determination sub-module, for being based on The multiple third subgraph determines target subgraph.

13. device according to claim 12, which is characterized in that the 4th determination sub-module is used for：

14. device according to claim 12, which is characterized in that the 5th determination sub-module is used for：