KR100810137B1 - Apparatus and method for reconstructing image using inverse discrete wavelet transforming - Google Patents

Abstract

A method and an apparatus for restoring an image by using inverse discrete wavelet transformation are provided to perform a two-dimensional inverse discrete wavelet conversion immediately when two-dimensional frequency data is inputted without latency. A frequency data reading unit(81) reads a plurality of secondary frequency data constituting a secondary frequency data block according to the predetermined order. The first inverse discrete wavelet conversion performing unit(82) calculates the first intermediate coefficient by performing inverse discrete wavelet conversion on the read secondary frequency data in the first direction, and restores primary frequency data by using the first intermediate coefficient. The second inverse discrete wavelet conversion performing unit(83) calculates the second intermediate coefficient by performing inverse discrete wavelet conversion on the primary frequency data in the second direction perpendicular to the first direction, and restores original pixel data by using the second intermediate coefficient.

Description

Apparatus and method for reconstructing image using inverse discrete wavelet transforming}

1 illustrates one embodiment of discrete wavelet transform;

2 illustrates one embodiment of inverse discrete wavelet transform;

3 is a diagram illustrating a discrete wavelet transform process using a line buffer memory proposed in the JPEG2000 standard specification.

4 illustrates an inverse discrete wavelet transform process using a line buffer memory.

5 is a diagram illustrating a lifting-based discrete wavelet transform process.

6 is a diagram illustrating a lifting-based discrete wavelet transform process in (9, 7) filter.

7 is a diagram illustrating a lifting-based inverse discrete wavelet transform process in a filter (9, 7) according to an embodiment of the present invention.

8 is a schematic block diagram of an image restoration apparatus using inverse discrete wavelet transform according to an embodiment of the present invention.

9 illustrates secondary frequency data according to an embodiment of the present invention.

10 is a diagram illustrating a first-order inverse discrete wavelet transform process in a row direction according to an embodiment of the present invention.

11 is a diagram illustrating a second inverse discrete wavelet transform process in a column direction according to an exemplary embodiment of the present invention.

<Description of the symbols for the main parts of the drawings>

80: image restoration device

81: frequency data reader

82: first inverse discrete wavelet transform performing unit

83: second inverse discrete wavelet transform performing unit

84: line memory

The present invention relates to image reconstruction, and more particularly, to an image reconstruction method and apparatus using lifting-based inverse discrete wavelet transform, which optimizes memory utilization.

In recent years, the demand for high quality image compression is increasing, and JPEG2000 (joint photographic experts group 2000) has been proposed as a new still image compression standard. Using this discrete wavelet transform (DWT), JPEG2000 provides a much higher compression rate than conventional JPEG based on discrete cosine transform (DCT). And the quality of the decompressed image is excellent.

However, the compression based on the discrete wavelet transform is more computational than the compression based on the discrete cosine transform, and there is a problem that a much larger memory must be used when performing the calculation. This is because the discrete cosine transform is applied to a block of pixels of 8x8 size, but the discrete wavelet transform should be applied to all pixels constituting the image. Inverse DWT, which reconstructs the original image from the data compressed by the discrete wavelet transform, also needs to be applied to the entire compressed data.

1 illustrates an embodiment of discrete wavelet transform. Here, L represents a low pass filtering (LPF) function, and H represents a high pass filtering (HPF) function.

Referring to FIG. 1, the discrete wavelet transform performs a two-dimensional operation.

First, the low pass filtering L ₁ (x) and the high pass filtering H ₁ (x) are performed in the horizontal direction on the original image data P 10. The first low frequency filtered image (L) 11 and the first high frequency filtered image (H) 16, which are half-processed by the down-sampling, are stored in the memory.

The low pass filtering image 11 and the first high pass filtering image 12 and the low pass filtering L ₂ (x) and the high pass filtering H ₂ (x) are performed in the vertical direction, respectively. . Second order low pass filtered images (LL (12), HL (17)) and second high pass filtered images (LH (13), HH (18)), which are half-processed by downsampling, are obtained. . That is, images of four frequency bands of LL, LH, HL, and HH are obtained.

Here, after the filtering in the horizontal direction, as many memory cells as the number of pixels of the original image data 10 are required, and as the size of the original image data 10 increases, the internal memory increases and the circuit size also increases significantly. .

2 illustrates an embodiment of inverse discrete wavelet transform. Where L represents an inverse low pass filtering (LPF) function and H represents an inverse high pass filtering (HPF) function.

Referring to FIG. 2, the inverse discrete wavelet transform performs a two-dimensional operation.

First, the second filtering image LL 21 and LH 22, HL 23 and HH (images for four frequency bands LL, LH, HL, and HH obtained through the discrete wavelet transform as described above) 24) upsample in the vertical direction. The image is doubled by padding the pixels of the second-filtered image. In addition, inverse low pass filtering (L ₃ (x)) and inverse high pass filtering (H ₃ (x)) are performed to generate the first filtered image (L (25), H (26)) and store in the memory.

In addition, up-sampling is performed in the horizontal direction on the first filtered images L (25) and H (26), and the image is doubled by padding the pixels of the first filtered image. Inverse low pass filtering (L ₄ (x)) and inverse high pass filtering (H ₄ (x)) are performed to obtain original image data (P) 27.

Here, the larger the size of the original image data 27, that is, the larger the size of the secondary filtering image (LL (21) and LH (22), HL (23) and HH (24)) increases the internal memory and the circuit The size is also greatly increased.

FIG. 3 is a diagram illustrating a discrete wavelet transform process using a line buffer memory proposed in the JPEG2000 standard specification to solve such a problem, and FIG. 4 is a diagram illustrating an inverse discrete wavelet transform process using a line buffer memory.

Referring to FIG. 3, in receiving the original image data 30, the line buffer memory 31 is stored therein to store image pixels of nine lines. That is, the line buffer memory 31 includes nine line memories 31-1, 31-2, 31-3, 31-4, 31-5, 31-6, 31-7, 31-8, 31-9. Each line memory (31-1, 31-2, 31-3, 31-4, 31-5, 31-6, 31-7, 31-8, or 31-9) consists of one line of image pixels Is saving. Line 9 is the number of lines required when using the (9, 7) filter in JPEG2000.

The data of the line buffer memory 31 stored by nine lines has the low-pass filtering and the high-pass filtering with nine pixel data 32 in the vertical direction instead of the horizontal direction. If this is performed nine times in the vertical direction, nine primary low frequency coefficients 33 and nine primary high frequency coefficients 34, which are filtered results, are generated.

In addition, when the low pass filtering and the high pass filtering are performed in the horizontal direction with respect to the nine primary low frequency coefficients 33 and the first high frequency coefficient 34, the second low frequency coefficients LL 35 and HL 37 Secondary high frequency coefficients LH 36 and HH 38 are obtained.

The above-described method does not require as much memory as the number of pixels of the original image data 30, but the minimum (9 in this example) line buffer memory capable of filtering in the horizontal direction after filtering in the vertical direction ( 31) to perform discrete wavelet transform. However, nine line memories are required to store nine lines of image pixel data, and the larger the original image data 30 becomes, the more a burden is placed on the circuit size.

Referring to FIG. 4, an inverse discrete wavelet transform process using a line buffer memory is illustrated. The input line memory 45 is 18 lines, but the second frequency coefficients LL 41, LH 42, HL 43, and HH 44 are reduced to 1/4 through down sampling. In the horizontal direction, it is reduced to half, which is equivalent to writing a 9 (= 18/2) line buffer memory. However, as mentioned above, even with this method, the line buffer memory that stores 9 lines of frequency data also becomes a burden on the circuit size as the image gets larger.

Accordingly, the present invention provides an image restoring method and apparatus using inverse discrete wavelet transform, which reduces the size of the memory used in the inverse discrete wavelet transform by 1/3 and economically reduces the production cost.

The present invention also provides a method and an apparatus for restoring an image using inverse discrete wavelet transform, which can perform two-dimensional inverse discrete wavelet transform as soon as secondary frequency data is inputted without delay.

In addition, the present invention provides a method and apparatus for image restoration using inverse discrete wavelet transform that can improve the amount of calculation and processing speed by applying a lifting-based wavelet transform method.

Other objects of the present invention will be readily understood through the following description.

According to an aspect of the present invention, in the apparatus for restoring an original image from secondary frequency data by performing inverse discrete wavelet transform, a plurality of secondary frequency data constituting a secondary frequency data block is read in a predetermined order. Coming frequency data reader; A first inverse discrete wavelet that inversely discrete-wavelets the second frequency data read by the frequency data reader in a first direction to calculate a first intermediate coefficient, and restores the first frequency data by using the first intermediate coefficient A conversion unit; And a second inverse discrete wavelet that transforms the first frequency data into a second discrete wavelet transform in a second direction perpendicular to the first direction to calculate a second intermediate coefficient and restore original pixel data using the second intermediate coefficient. An image reconstruction device using inverse discrete wavelet transform is provided.

Preferably, the first direction may be any one of a row direction or a column direction, and the second direction may be the other of the row direction or the column direction.

According to another aspect of the present invention, the frequency data reading unit may include a second frequency data block including a low low filtering coefficient (LL), a low high filtering coefficient (LH), a high low filtering coefficient (HL), and a high high filtering coefficient (HH). The second frequency data block may include one or more second frequency data small blocks in order in the first direction with respect to each line in the second direction.

According to another aspect of the present invention, the first frequency data includes first low frequency data and first high frequency data, and the first inverse discrete wavelet transform performing unit comprises the low low filtering coefficient LL and the low high filtering coefficient. Inverse discrete wavelet transform (LH) to restore the first low frequency data, and inverse discrete wavelet transform the high and low filtering coefficients (HL) and the high filtering coefficient (HH) to restore the first high frequency data. An image restoration apparatus using inverse discrete wavelet transform may be provided.

According to still another aspect of the present invention, the apparatus may further include a line memory configured to store the first high frequency data, the first low frequency data, and the first intermediate coefficient, respectively.

According to another aspect of the present invention, when the secondary frequency data block is a secondary frequency data block except for the first secondary frequency data block in either of the first direction or the second direction. The first inverse discrete wavelet transform performing unit may include first frequency data and first intermediate coefficients of a previous secondary frequency data block read from the line memory, low height filtering coefficients (LH), and high height filtering of the previous secondary frequency data block. An apparatus for restoring an image using inverse discrete wavelet transform may be provided to perform inverse discrete wavelet transform on the coefficient HH and the second frequency data read by the frequency data reader.

 According to yet another aspect of the present invention, an inverse discrete wavelet transform further comprises a register for storing the first high frequency data, the second intermediate coefficients, and the restored original pixel data of a previous second frequency data block. The image restoration apparatus used may be provided.

In addition, when the secondary frequency data block is a frequency data block except for the first secondary frequency data block in one of the first direction and the second direction, the second inverse discrete wavelet transform performing unit may include the register. The first high frequency data of the previous frequency data block read from the second first coefficient, the second intermediate coefficient and the restored original pixel data and the first low frequency data read from the frequency data reader, and the first high frequency data. The second intermediate coefficients may be calculated by inverse discrete wavelet transformation in two directions, and the plurality of original pixel data may be restored.

Preferably, the first inverse discrete wavelet transform performing unit may perform a lifting based inverse discrete wavelet transform.

Also preferably, the second inverse discrete wavelet transform performing unit may perform a lifting based inverse discrete wavelet transform.

According to another aspect of the present invention, in the apparatus for reconstructing the original image from the secondary frequency data by performing inverse discrete wavelet transform, (a) a plurality of secondary frequencies constituting a frequency data block in a predetermined order Reading data; (b) inverse discrete wavelet transforming the second frequency data in a first direction to calculate a first intermediate coefficient and reconstructing the primary frequency data using the first intermediate coefficient; And (c) inverse discrete wavelet transforming the first frequency data in a second direction perpendicular to the first direction to calculate a second intermediate coefficient and reconstruct original pixel data using the second intermediate coefficient. An image reconstruction method using inverse discrete wavelet transform may be provided.

Preferably, the first direction may be either a row direction or a column direction, and the second direction may be the other of the row direction or the column direction.

In addition, step (a) includes a second frequency data block including a low low filtering coefficient (LL), a low high filtering coefficient (LH), a high low filtering coefficient (HL), and a high high filtering coefficient (HH) on each line in the second direction. Can be read in order in the first direction.

According to another aspect of the present invention, the first frequency data includes first order low frequency data and first order high frequency data, and the step (b) includes the low low filtering coefficient LL and the low high filtering coefficient LH. The first low frequency data may be restored by performing inverse discrete wavelet transform, and the first high frequency data may be restored by performing inverse discrete wavelet transformation of the high and low filtering coefficients HL and HH.

According to still another aspect of the present invention, the method may further include storing the first high frequency data, the first low frequency data, and the first intermediate coefficient in a line memory, respectively.

According to another aspect of the present invention, when the frequency data block is a frequency data block except for the first frequency data block in either of the first direction or the second direction, step (b) Low frequency filtering coefficient (LH) and high height filtering coefficient (HH) and the frequency data dock of the first frequency data, the first intermediate coefficient of the previous frequency data block read from the line memory and the previous frequency data block read from the line memory The secondary frequency data read by the output unit may be subjected to inverse discrete wavelet transform in a first direction to calculate a first intermediate coefficient and restore the primary frequency data.

Also preferably, the step (c) may inverse discrete wavelet transform the first low frequency data and the first high frequency data to calculate a second intermediate coefficient and restore the original pixel data.

According to another aspect of the present invention, the method may further include storing the first high frequency data of the previous frequency data block, the second intermediate coefficient, and the restored original pixel data in a register.

According to another aspect of the present invention, when the frequency data block is a frequency data block except for the first frequency data block in either of the first direction or the second direction, step (c) The first high frequency data of the previous frequency data block read from the register, the second intermediate coefficients and the restored original pixel data and the first low frequency data and the first high frequency data read from the frequency data reader. The second intermediate coefficients may be calculated by inverse discrete wavelet transformation in two directions, and the plurality of original pixel data may be restored.

Preferably, the step (b) may be provided with an image reconstruction method using the inverse discrete wavelet transform, characterized in that for performing the lifting-based inverse discrete wavelet transform.

Also preferably, the step (c) may be provided with an image reconstruction method using the inverse discrete wavelet transform, characterized in that for performing the lifting-based inverse discrete wavelet transform.

Hereinafter, with reference to the accompanying drawings will be described in detail a preferred embodiment of the image reconstruction method and apparatus using inverse discrete wavelet transform according to the present invention. In describing the present invention, when it is determined that the detailed description of the related known technology may unnecessarily obscure the subject matter of the present invention, the detailed description thereof will be omitted. Numbers (eg, first, second, etc.) used in the description of the present specification are merely identification symbols for sequentially distinguishing identical or similar entities.

In the present invention, an image reconstruction method using inverse discrete wavelet transform and a device thereof will be described based on the reconstruction of original pixel data from a first vertical line to a horizontal direction in two-dimensional image data. However, it is of course possible to restore the original pixel data in the vertical direction from the first horizontal line.

Prior to describing the inverse discrete wavelet transform method of the image data, a lifting based discrete wavelet transform method will be described.

FIG. 5 is a diagram illustrating a lifting-based discrete wavelet transform process, and FIG. 6 is a diagram illustrating a lifting-based discrete wavelet transform process in the filter (9, 7).

Lifting-based discrete wavelet transform is an efficient filtering method that includes down-sampling as well as low-pass filtering and high-pass filtering that were separately calculated in the conventional discrete wavelet transform.

Referring to FIG. 5, the lifting transformation process includes a lazy transform part 51, a predictor 52, an updater 53, and an output unit 54, 55.

The lazy conversion unit 51 compares whether the index of the incoming input data is odd or even and sends the odd and even numbers separately. The prediction unit P (z) 52 is a portion that multiplies a prediction coefficient by an input value, and the updater U (z) 53 is a portion that multiplies an input value by an update coefficient. The output parts K ₀ , K ₁ ; 54, 55 are parts for normalizing the calculated values. The converted value is shown in Equation 1 below.

In Equation 1, when N is 1, it becomes a (5, 3) filter, and when N is 2, it becomes a (9, 7) filter. When expressing the filter, the preceding number represents the number of taps of the low frequency filter, and the latter number represents the number of taps of the high frequency filter. For the (9, 7) filter, the low frequency filter has 9 taps and the high frequency filter has 7 taps.

A discrete wavelet transform using the (9, 7) filter will be described with reference to FIG.

Referring to the first dotted line 60 of FIG. 6, it can be seen that one low frequency data and one high frequency data are obtained using five pixel data of the image data. The second dotted line 61 shows that three pixel data and high frequency data and intermediate coefficients of the previous data block are needed to acquire the low frequency data and the high frequency data of the next pixel block.

7 is a diagram illustrating a lifting-based inverse discrete wavelet transform process in a filter (9,7) according to an embodiment of the present invention.

Referring to FIG. 7, a process of calculating reconstruction data from frequency data is shown. Here, each data is as follows.

s ₀ ¹ , s ₁ ¹ : Low frequency data of the 1st frequency data group

d ₀ ¹ , d ₁ ¹ : High frequency data of the 1st frequency data group

s _i-1 ¹ : Low-frequency data of the i-th frequency data group

d _i-1 ¹ : High frequency data of the i-th frequency data group

s ₀ ⁰ : Reconstruction data of the first book circle data group

s _i-1 ⁰ : First restore data of the i th restore data group

d _i-2 ⁰ : Second restoration data of the i th restoration data group

(i≥2))

The lifting based inverse discrete wavelet transform process is similar to the reverse arrangement of the lifting based discrete wavelet transform process shown in FIG. Coefficients multiplied in the discrete wavelet transform process are converted into positive numbers and negative numbers.

In the top row of Fig. 7, frequency data groups are arranged in order. The low frequency coefficients and the high frequency coefficients are alternately arranged, and one low frequency coefficient and one high frequency coefficient form one frequency data group.

When inverse discrete wavelet transform is performed on the first frequency data group s ₀ ¹ , d ₀ ¹ , s ₁ ¹ , d ₁ ¹ , one piece of reconstructed data is obtained. In this process, intermediate coefficients I ₀ and I ₁ are _first calculated from s ₀ ¹ and d ₀ ¹ . And I ₂ is calculated from d ₀ ¹ and I ₀ and I ₁ . Finally, s ₀ ⁰ is obtained from I ₀ and I ₂ . The obtained reconstruction data is s ₀ ^{0, which} is the first reconstruction data of the first reconstruction data group. If the above process is summarized, the following equation may be obtained.

I ₀ = s ₀ ¹ + u ₂ d ₀ ¹

I ₁ = s ₁ ¹ + u ₂ (d ₀ ¹ + d ₁ ¹ )

I ₂ = p ₂ (I ₀ + I ₁ ) + d ₀ ¹

s ₀ ⁰ = I ₀ + u ₁ I ₂

When inverse discrete wavelet transform is performed on the second and subsequent i th frequency data groups s _i ¹ and d _i ¹ , two pieces of reconstructed data s _i-1 ⁰ and d _i-2 ⁰ are obtained. do. In order to acquire two pieces of reconstruction data, the high frequency coefficient d _i-1 ¹ of the previous frequency data group, the reconstruction data s _i-1 ⁰ obtained from the previous reconstruction data group, and the previous secondary frequency data group I ₁ and I ₂ calculated during the data restoration process are needed.

The second frequency data group is described as follows.

To obtain the reconstructed data d ₀ ⁰ and s ₁ ⁰ , first, the second frequency data s ₂ ¹ and d ₂ ¹ of the _second frequency data group are read and the d ₁ ¹ of the previous frequency data group is read. First, the intermediate coefficient I ₁₁ is calculated using d ₁ ¹ and s ₂ ¹ and d ₂ ¹ . Then, an intermediate coefficient I ₂₂ is calculated from the intermediate coefficients I ₁ and I ₁₁ and d ₁ ¹ , which are already calculated in the previous secondary frequency data block.

Then, reconstruction data s ₁ ⁰ is obtained from I ₂ and I ₁ , which are intermediate coefficients already calculated in I ₂₂ and the previous secondary frequency data block. And d ₀ ⁰ can be restored by calculating using s ₀ ⁰ , I ₂ and s ₁ ⁰ , which are reconstructed data recovered from the previous secondary frequency data group. As a result, it can be seen that image data s ₁ ⁰ and d ₀ ⁰ are reconstructed through the first and second inverse discrete wavelet transforms in the second secondary frequency data group. The above-described process is represented by the following equation.

I ₁₁ = s ₂ ¹ + u ₂ (d ₁ ¹ + d ₂ ¹ )

I ₂₂ = p ₂ (I ₁ + I ₁₁ ) + d ₁ ¹

s ₁ ⁰ = I ₁ + u _{1 (} I ₂ + I ₂₂ )

d ₀ ⁰ = I ₂ + p _{1 (} s ₀ ⁰ + s ₁ ⁰ )

By performing inverse discrete wavelet transform according to the above-described process, reconstructed data can be obtained simply and efficiently.

The original image data may be reconstructed by performing the inverse discrete wavelet transform of the second frequency data obtained through the two discrete wavelet transforms once in the column direction and the row direction, respectively.

In the following description, the intermediate coefficients calculated in performing the first inverse discrete wavelet transform are determined by performing the first intermediate coefficient and the second inverse discrete wavelet transform regardless of the number of secondary frequency data blocks on which the inverse discrete wavelet transform is performed. The intermediate coefficient computed is called a 2nd intermediate coefficient. Then, the first intermediate coefficients I ₁ ′ and I ₂ ′ and the second intermediate coefficients are unified with I ₁ ″ and I ₂ ″.

An image reconstruction apparatus using inverse discrete wavelet transform for reconstructing original image data by using the above-described lifting-based inverse discrete wavelet transform method will be described below.

FIG. 8 is a schematic block diagram of an image reconstruction apparatus using inverse discrete wavelet transform according to an embodiment of the present invention, FIG. 9 is a view showing secondary frequency data according to an embodiment of the present invention. 10 is a diagram illustrating a first inverse discrete wavelet transform process in a row direction according to an exemplary embodiment of the present invention, and FIG. 11 is a second inverse discrete wavelet transform in a column direction according to an exemplary embodiment of the present invention. Figure showing the process.

Referring to FIG. 8, the image reconstruction apparatus 80 using the inverse discrete wavelet transform includes a frequency data reader 81, a first inverse discrete wavelet transform performer 82, and a second inverse discrete wavelet transform performer 83. It includes. The image reconstruction apparatus 80 using inverse discrete wavelet transform may further include three line memories 84. Hereinafter, the first inverse discrete wavelet transform performer 82 performs inverse discrete wavelet transform in a row direction, and the second inverse discrete wavelet transform performer 83 inverse discrete wavelet transform in a column direction It is assumed that the following is performed, but it may be reversed. Hereinafter, the inverse discrete wavelet transform by the first inverse discrete wavelet transform performer 82 is referred to as a first order inverse discrete wavelet transform, and the inverse discrete wavelet transform by the second inverse discrete wavelet transform performer 83 is secondary. This is called an inverse discrete wavelet transform.

In addition, below, the first inverse discrete wavelet transform is the row direction and the second inverse discrete wavelet transform is the column direction for each pixel data of the original image, but the first inverse discrete wavelet transform is the column direction, It is obvious that the same applies to the case where the second-order inverse discrete wavelet transform is in the row direction.

The frequency data reading unit 81 performs a predetermined sequence of the plurality of secondary frequency data generated by performing the discrete wavelet transform twice (the first discrete wavelet transform in the row direction and the second discrete wavelet transform in the column direction). Secondary frequency data block). The plurality of secondary frequency data may be divided into secondary frequency data blocks and read out in a specific order by the frequency data reader 81, which will be described later with reference to FIG. 9.

The first inverse discrete wavelet transform performing unit 82 performs first inverse discrete wavelet transform in the row direction on the second frequency data read by the frequency data reading unit 81, thereby performing first intermediate coefficients I ₁ ′ and I ₂ ′. ), And obtain primary frequency data (primary low frequency data and primary high frequency data).

The second inverse discrete wavelet transform performing unit 83 performs second inverse discrete wavelet transform in the column direction on the first frequency data acquired by the first inverse discrete wavelet transform performing unit 82 to generate the second intermediate coefficient I ₁ ″. And I ₂ ″), to obtain original pixel data.

In acquiring the primary frequency data in the row direction, in order to acquire the primary frequency data after the first primary frequency data, the first inverse discrete wavelet transform performing unit 82 performs frequency data in the current secondary frequency data block. In addition, it requires secondary high frequency data in the previous secondary frequency data block and previously obtained primary frequency data and a first intermediate coefficient.

Thus, the primary frequency data (ie, primary low frequency data and primary high frequency data) and the first intermediate coefficients are stored in the line memory 84. In addition, when the first inverse discrete wavelet transform performer 82 performs inverse discrete wavelet transform on the second order frequency data block, the first inverse discrete wavelet transform performer 82 may include the second frequency data in the previous second frequency block. The primary frequency data stored in the line memory 84 and the first intermediate coefficients of the previous secondary frequency data block are used.

After performing the first inverse discrete wavelet transform in the row direction, in performing the second inverse discrete wavelet transform in the column direction, the second inverse discrete wavelet transform undergoes a process similar to the first inverse discrete wavelet transform.

In acquiring original pixel data in the column direction, in order to acquire original pixel data after the first original pixel data, the second inverse discrete wavelet transform performing unit 83 transfers in addition to the primary frequency data in the current primary frequency data group. First order high frequency data in the primary frequency data group, second intermediate coefficients and previously acquired original pixel data are required.

Accordingly, the first high frequency data, the second intermediate coefficients, and the original pixel data in the previous first frequency data group are respectively stored in a register, and the second inverse discrete wavelet transform performing unit 83 performs the first frequency in the following order. When performing inverse discrete wavelet transform on a data group, the primary high frequency data, the second intermediate coefficients, and the original pixel data in the previous primary frequency data group stored in the register may be used.

Hereinafter, a process of obtaining original pixel data by converting secondary frequency data in the image recovery apparatus 80 using inverse discrete wavelet transform will be described with reference to FIGS. 9 through 11.

It is assumed that the secondary frequency data are arranged as shown in FIG. The frequency data small block means a group of frequencies in which m and n each have the same value. L (m, n) and H (m, n) form one primary frequency data small block, LL (m, n), LH (m, n), HL (m, n), HH (m , n) form one secondary frequency data small block (where m ≧ 1, n ≧ 1). Secondary frequency data small blocks of LL (m, n), LH (m, n), HL (m, n), and HH (m, n) are located at coordinates (m, n) among the secondary frequency data. Set to. Where m is the row coordinate of the secondary frequency data small block, and n is the column coordinate of the secondary frequency data small block.

The secondary frequency data block consists of one or more secondary frequency data small blocks.

In the present specification, the secondary frequency data block refers to 90 (11), 90 (12), 90 (13), 90 (21), and the like of FIG. 9. Referring to 90 (11), 90 (12), 90 (13), 90 (21), and 90 (22) of FIG. 9, the same secondary frequency data block may have different sizes. That is, 90 (11) is composed of four secondary frequency data small blocks, and 90 (12), 90 (13), and 90 (21) are two secondary frequency small blocks, 90 (22) and 90 (23). , 90 (24), 90 (32), etc., consist of only one secondary frequency data small block. In the case of 90 (11), in the inverse discrete wavelet transform using the (9,7) filter, 90 (11) in which the first inverse discrete wavelet transform is performed need four quadrature frequency data, each of two in width and length. . In the next secondary frequency data block, the discrete wavelet transform is performed using the result calculated from the previous frequency data block (first intermediate coefficient, previous secondary frequency data of the previous frequency data block, etc.). That is, if the first secondary frequency data block is not the first secondary frequency data block in the row direction or the column direction, the operation is performed using the result of the inverse discrete wavelet transform of the previous secondary frequency data block. Also, the number of secondary frequency data small blocks constituting the secondary frequency data block is also reduced.

Among the rows or columns of the secondary frequency data blocks, the secondary frequency data blocks located in the first row or the first column correspond to the first secondary frequency data block in the row direction or the column direction. That is, if the secondary frequency data block is represented by 90 (pq) in which p is an order of rows and q is an order of columns in FIG. 9, if at least one of p or q is 1, the first 2 in the row direction or column direction Corresponds to a differential frequency data block (eg, 90 (12), 90 (13), 90 (21), etc.). Further, a secondary frequency data block in which both p and q are 2 or more becomes a frequency data block except for the first secondary frequency data block in either the first direction or the second direction (eg, 90 (22)). , 90 (23), 90 (24), etc.).

And "previous frequency data block in row direction " may be described as follows. For example, at 90 (12), the previous frequency data block in the row direction becomes 90 (11), and in 90 (23), the previous frequency data block in the row direction becomes 90 (22). In the same principle, the data block in the column direction at 90 (21) becomes 90 (11).

In order to reconstruct the original image in the row direction, the frequency data reader 81 has four secondary frequency data sources located at (1,1), (1,2), (2,1) and (2,2). From a first secondary frequency data block 90 (11) consisting of blocks, consisting of two secondary frequency data small blocks located at (1,3) and (2,3) in the row direction Third secondary frequency data block 90 consisting of a second secondary frequency data block 90 (12) and two secondary frequency data small blocks located at (1,4) and (2,4). The data is read in the order of (13). When reconstructing the original image in the column direction, the frequency data reader 81 reads data of the secondary frequency data blocks 90 (11), 90 (21), and 90 (31) in the column direction.

Hereinafter, it is assumed that the original image is restored in the order of the second row and the third row after the original image is restored in the row direction with respect to the first row. Hereinafter, the original image may be restored in the order of the second column and the third column after the original image is restored in the column direction with respect to the first column.

The secondary frequency data in the secondary frequency data block 90 (11) located in the first column (located at (1,1)) in the first row in the row direction is LL (1,1), LH (1,1). ), LL (1,2), LH (1,2), HL (1,1), HH (1,1), HL (1,2), HH (1,2), LL (2,1) , LH (2,1), LL (2,2), LH (2,2), HL (2,1), HH (2,1), HL (2,2), HH (2,2) . Here, the secondary frequency data is composed of coefficients generated by performing the above-described two discrete wavelet transforms. A total of four coefficients are generated according to the filtering method in the first discrete wavelet transform and the filtering method in the second discrete wavelet transform among two discrete wavelet transforms. The low frequency filtered coefficients during the first discrete wavelet transform are divided into the low low filtering coefficients LL filtered by the low frequency during the second or more wavelet transform, and the low high filtering coefficients LH filtered by the high frequency. The high frequency filtered coefficients in the first discrete wavelet transform are divided into the high and low filtering coefficients HL filtered low frequency in the second discrete wavelet transform and the high and high filtering coefficients HH filtered high frequency.

Referring to Figure 10, the first secondary frequency data block 90 (11) is the first frequency of LL (1, 1) and LH (1, 1), LL (1, 2) and LH (1, 2) The data group 101 is divided into HL (1,1) and HH (1,1), HL (1,2) and HH (1,2) into a second frequency data group 102. Further, LL (2,1) and LH (2,1), LL (2,2) and LH (2,2) are the third frequency data group 103, and HL (2,1) and HH (2). HL (2,2) and HH (2,2) are divided into a fourth frequency data group 104. In addition, the first inverse discrete wavelet transform performing unit 82 performs first-order low frequency, which is reconstruction data, by performing a lifting-based inverse discrete wavelet transform on the second frequency data in the first and third frequency data groups 101 and 103. The data L (1,1) and L (2,1) are restored. The first inverse discrete wavelet transform performing unit 82 also performs lifting based inverse discrete wavelet transform on the second frequency data in the second and fourth frequency data groups 102 and 104, thereby reconstructing the first-order high frequency wave. Data H (1,1) and H (2,1) are respectively restored.

The first primary frequency data group 100 (11) is composed of two primary low frequency data and two primary high frequency data (L (1,1), H (1,1), L (2,1), H (2, 1)). The first primary frequency data group 100 (11) is needed to recover the next second and third primary frequency data groups 100 (12) and 100 (13), and thus to the line memory 84. Stored.

The first intermediate coefficients are calculated during the first primary frequency data restoration process, and are represented by I ₁ ′ and I ₂ ′. The first intermediate coefficient is also stored in the line memory 84 because it is needed for the next primary frequency data calculation.

The first inverse discrete wavelet transform performing unit 82 performs an analysis of the first secondary frequency data block 90 (11) that has been read for inverse discrete wavelet transform of the second secondary frequency data block 90 (12) in the row direction. The second discrete wavelet transform of the second frequency data requires a high frequency filtering group consisting of high frequency filtered LH (1,2), HH (1,2), LH (2,2), and HH (2,2). In addition, the first primary frequency data group 100 (11) obtained previously and two first intermediate coefficients I ₁ ′, I ₂ ′ calculated during the first primary frequency data acquisition process are required. In addition, the first inverse discrete wavelet transform performing unit 82 also needs the first frequency data 100 (11) obtained by the inverse discrete wavelet transform of the first secondary frequency data block 90 (11). The first intermediate coefficients I ₁ ′, I ₂ ′ and primary frequency data 100 (11) are read from the line memory 84.

The first inverse discrete wavelet transform performer 82 may include previously obtained first primary frequency data, two first intermediate coefficients I ₁ ′, I ₂ ′ calculated during a first primary frequency data acquisition process, Secondary frequency data LH (1,2) and HH (1,2), LH (2,2) and HH (2,2) of the secondary frequency data block 90 (11) and the secondary frequency data block The inverse discrete wavelet transform described above is performed on the secondary frequency data of 90 (12).

The first low frequency data L (1,2), L (2,2), L (1,3), L (2,3) and primary high frequency data H (1,2), H (2) 2, H (1,3), and H (2,3). The primary frequency data includes a second primary frequency data group 100 (12) consisting of L (1,2) and H (1,2), L (2,2) and H (2,2); It is grouped into a third primary frequency data group 100 (13) consisting of L (1,3) and H (1,3), L (2,3) and H (2,3).

The third primary frequency data group 100 (13) of the primary frequency data group is needed to obtain the next original pixel data, and is thus stored in the line memory 84.

Thereafter, the first inverse discrete wavelet transform performer 82 reads the second secondary frequency data block 90 (12) already read for the inverse discrete wavelet transform of the third secondary frequency data block 90 (13) in the row direction. A second high frequency filtering group 96 consisting of LH (0,2) and HH (0,2), LH (1,2) and HH (1,2) is required for the second discrete wavelet transform of Shall be. In addition, the first inverse discrete wavelet transform performing unit 82 may perform a first intermediate frequency data 100 (13) and a first intermediate frequency obtained by the inverse discrete wavelet transform of the second secondary frequency data block 90 (12). The coefficients I ₁ ′, I ₂ ′ are also required, so that the third primary frequency data 100 (13) and the first intermediate coefficients are read from the line memory 84 to perform a first inverse discrete wavelet transform. do.

Thereafter, it is possible to restore the primary frequency data by performing inverse discrete wavelet transform in the same manner as the second and third secondary frequency data blocks as described above with respect to the fourth and fifth secondary frequency data blocks in the row direction. .

Referring to FIG. 11, the second inverse discrete wavelet transform performer 83 performs second inverse discrete wavelet transform in the column direction on the restored first frequency data after the first inverse discrete wavelet transform is performed in the row direction. Perform. In the first secondary frequency data block 90 (11), LL (1,1) and LH (1,1), LL (1,2) and LH (1,2) are transferred to the first frequency data group 101. , HL (1,1) and HH (1,1), HL (1,2) and HH (1,2) are divided into a second frequency data group 102. LL (1,0) and LH (1,0), LL (1,1) and LH (1,1) are the third frequency data group 103, and HL (2,1) and HH (2, 1), HL (2,2) and HH (2,2) are divided into a fourth frequency data group 104. The first inverse discrete wavelet transform performing unit 82 performs first-order inverse discrete wavelet transform in the row direction to perform the first low frequency data L (1,1), L (2,1) and the first high frequency data H (1, 1), H (2,1) is restored. In the process, second intermediate coefficients I ₁ "and I ₂ " are calculated.

The second inverse discrete wavelet transform performer 83 includes a first low frequency data L (1,1), L (2,1) and first high frequency data H (1,1) and H (2,1). The first original pixel data P (1,1) 110 (11) is reconstructed by performing the inverse discrete wavelet transform based on the first primary frequency data group 100 (11) in the column direction.

In the second secondary frequency data block 90 (21) in the column direction, LL (3,1) and LH (3,1), LL (3,2) and LH (3,2) represent a fifth frequency data group ( 105), HL (3,1) and HH (3,1), HL (3,2) and HH (3,2) are divided into a sixth frequency data group 106. The first inverse discrete wavelet transform performing unit 82 performs the first inverse discrete wavelet transform in the row direction to restore the first low frequency data L (3,1) and the first high frequency data H (3,1).

The second inverse discrete wavelet transform performing unit 83 performs the second inverse discrete wavelet transform in the column direction on the first low frequency data L (3,1) and the first high frequency data H (3,1).

The second inverse discrete wavelet transform performing unit 83 performs first original pixel data P (1,1) 110 on the second inverse discrete wavelet transform to perform the second inverse discrete wavelet transform in the column direction. (11)), the first high frequency data H (2,1) in the first primary frequency data group 100 (11) and the second intermediate coefficients I ₁ ″, I ₂ ″ are required.

The first original pixel data P (1,1) 110 (11), the first order high frequency data H (2,1) and the second intermediate coefficients I ₁ ", I ₂ " are each temporarily stored in a register There may be.

The second inverse discrete wavelet transform performing unit 83 includes the first original pixel data P (1,1) and the first high frequency data H (2,1), the first low frequency data L (3,1) and the first high frequency. Two original pixel data P (2,1) 110 (21) and P (3,1) 110 (31) are restored using the data H (3,1).

Also, in the third secondary frequency data block 90 (31) in the column direction, LL (4,1) and LH (4,1), LL (4,2) and LH (4,2) are the seventh frequency data group. At 107, HL (4,1) and HH (4,1), HL (4,2) and HH (4,2) are divided into eighth frequency data group 108. The first inverse discrete wavelet transform performer 82 performs the first inverse discrete wavelet transform in the row direction to restore the first low frequency data L (4,1) and the first high frequency data H (4,1).

The second inverse discrete wavelet transform performing unit 83 performs the third original pixel data P reconstructed by the previous second inverse discrete wavelet transform in order to perform the second inverse discrete wavelet transform in the column direction according to Equation 2 described above. 3,1) 110 (31), second intermediate coefficients I ₁ ″, I ₂ ″ and primary high frequency data H (3, 1).

The third original pixel data P (3,1), the second intermediate coefficients I ₁ ″, I ₂ ″ and the primary high frequency data H (3, 1) may be temporarily stored in a register.

The second inverse discrete wavelet transform performing unit 83 performs the third original pixel data P (3,1) 110 (31), the second intermediate coefficients I ₁ ″, I ₂ ″ and the primary high frequency data H ( 2, original pixel data P (4,1) and P (5,1) are recovered using the first low frequency data L (4,1) and the first high frequency data H (4,1). do.

As described above with respect to the fourth and fifth secondary frequency data blocks in the column direction, it is possible to restore the original pixel data by performing the second inverse discrete wavelet transform in the same manner as the second and third secondary frequency data blocks. Although the first inverse discrete wavelet transform is a row direction and the second inverse discrete wavelet transform is a process of restoring original pixel data centering on a column direction, the first inverse discrete wavelet transform is a column direction and a second order. Obviously, the inverse discrete wavelet transform is applicable even in the row direction.

As described above, the image reconstruction method and the apparatus using the inverse discrete wavelet transformation according to the present invention has the effect of reducing the size of the memory used in the inverse discrete wavelet transformation by 1/3 and economically reducing the production cost. have.

In addition, two-dimensional inverse discrete wavelet transform may be performed immediately after the second frequency data is input without any latency.

In addition, the calculation and the processing speed can be improved by applying the lifting-based wavelet transform method.

Although the above has been described with reference to a preferred embodiment of the present invention, those skilled in the art to which the present invention pertains without departing from the spirit and scope of the present invention as set forth in the claims below It will be appreciated that modifications and variations can be made.

Claims (21)

An apparatus for reconstructing an original image from secondary frequency data by performing inverse discrete wavelet transform, A frequency data reader for reading a plurality of secondary frequency data constituting the secondary frequency data block in a predetermined order; A first inverse discrete wavelet that inversely discrete-wavelets the second frequency data read by the frequency data reader in a first direction to calculate a first intermediate coefficient, and restores the first frequency data by using the first intermediate coefficient A conversion unit; And Inverse discrete wavelet transform the first frequency data in a second direction perpendicular to the first direction to calculate a second intermediate coefficient and use the second intermediate coefficient to reconstruct original pixel data. An image reconstruction apparatus using inverse discrete wavelet transform including an execution unit. The method of claim 1, And wherein the first direction is any one of a row direction or a column direction, and the second direction is the other one of a row direction or a column direction. The method of claim 1, The frequency data readout unit includes a second frequency data block including a low low filtering coefficient LL, a low high filtering coefficient LH, a high low filtering coefficient HL, and a high high filtering coefficient HH for each line in the second direction. To read in the order in the first direction characterized in that, And said secondary frequency data block comprises one or more said secondary frequency data small blocks. The method of claim 3, The primary frequency data includes primary low frequency data and primary high frequency data, The first inverse discrete wavelet transform performing unit performs inverse discrete wavelet transform on the low low filtering coefficient LL and the low high filtering coefficient LH to restore the first low frequency data, and the high low filtering coefficient HL and the high height filtering. And an inverse discrete wavelet transform to inverse discrete wavelet transform the coefficient (HH). The method of claim 4, wherein And a line memory for storing the first high frequency data, the first low frequency data, and the first intermediate coefficient, respectively. The method of claim 5, When the secondary frequency data block is a secondary frequency data block except for the first secondary frequency data block in either of the first direction or the second direction The first inverse discrete wavelet transform performing unit may include first frequency data and first intermediate coefficients of a previous secondary frequency data block read from the line memory, low height filtering coefficients (LH), and high height filtering of the previous secondary frequency data block. And an inverse discrete wavelet transform of the coefficient (HH) and the second frequency data read by the frequency data reader in the first direction. The method of claim 4, wherein And a register for storing the first high frequency data, the second intermediate coefficients, and the reconstructed original pixel data of a previous second frequency data block, respectively. The method of claim 7, wherein When the secondary frequency data block is a frequency data block except for the first secondary frequency data block in either of the first direction or the second direction The second inverse discrete wavelet transform performing unit includes the first high frequency data of a previous frequency data block read from the register, the second intermediate coefficient and the restored original pixel data, and the first data read from the frequency data reader. And an inverse discrete wavelet transform of the low frequency data and the first high frequency data in a second direction to calculate a second intermediate coefficient, and reconstruct a plurality of original pixel data. Claim 9 was abandoned upon payment of a set-up fee. The method of claim 1, And the first inverse discrete wavelet transform performing unit performs a lifting based inverse discrete wavelet transform. Claim 10 was abandoned upon payment of a setup registration fee. The method of claim 1, The second inverse discrete wavelet transform performing unit performs an inverse discrete wavelet transform based on a lifting based inverse discrete wavelet transform. An apparatus for reconstructing an original image from secondary frequency data by performing inverse discrete wavelet transform, (a) reading a plurality of secondary frequency data constituting a frequency data block in a predetermined order; (b) inverse discrete wavelet transforming the second frequency data in a first direction to calculate a first intermediate coefficient and reconstructing the primary frequency data using the first intermediate coefficient; And (c) inverse discrete wavelet transforming the first frequency data in a second direction perpendicular to the first direction to calculate a second intermediate coefficient and reconstructing original pixel data using the second intermediate coefficient. An image reconstruction method using inverse discrete wavelet transform. Claim 12 was abandoned upon payment of a registration fee. The method of claim 11, And the first direction is one of a row direction and a column direction, and the second direction is the other one of the row direction and the column direction. The method of claim 12, Repeat steps (a) to (c), In step (a), a second frequency data block comprising a low low filtering coefficient (LL), a low high filtering coefficient (LH), a high low filtering coefficient (HL), and a high high filtering coefficient (HH) is applied to each line in the second direction. The image reconstruction method using inverse discrete wavelet transform, characterized in that the reading in the first direction in order. The method of claim 13, The primary frequency data includes primary low frequency data and primary high frequency data, In the step (b), the low-low filtering coefficient LL and the low-high filtering coefficient LH are inverse discrete wavelet transformed to restore the first low-frequency data, and the high-low filtering coefficient HL and the high-high filtering coefficient HH. ) Is used to inverse discrete wavelet transform to restore the first high frequency data. The method of claim 14, (d) storing the first high frequency data, the first low frequency data and the first intermediate coefficient in a line memory, respectively; And repeating the steps (a) to (d). The method of claim 15, When the frequency data block is a frequency data block except for the first frequency data block in one of the first direction and the second direction. The step (b) includes the low frequency filtering coefficient (LH) and the high height filtering coefficient (HH) of the first frequency data, the first intermediate coefficient of the previous frequency data block read from the line memory, and the previous frequency data block read from the line memory. And the inverse discrete wavelet transform of the second frequency data read by the frequency data reader in a first direction to calculate a first intermediate coefficient, and to restore the first frequency data using the inverse discrete wavelet transform. How to restore an image. Claim 17 was abandoned upon payment of a registration fee. The method of claim 14, In the step (c), the inverse discrete wavelet transform of the first low frequency data and the first high frequency data is performed to calculate a second intermediate coefficient, and the original pixel data is restored. Way. Claim 18 was abandoned upon payment of a set-up fee. The method of claim 17, (f) storing the first high frequency data of the previous frequency data block, the second intermediate coefficients and the recovered original pixel data in a register, And repeating the steps (a) to (f). Claim 19 was abandoned upon payment of a registration fee. The method of claim 18, When the frequency data block is a frequency data block except for the first frequency data block in either of the first direction or the second direction The step (c) comprises the primary high frequency data of the previous frequency data block read from the register, the second intermediate coefficient and the restored original pixel data and the primary low frequency data read from the frequency data reader, A method of restoring an image using inverse discrete wavelet transform, wherein the first high frequency data is inverse discrete wavelet transformed in a second direction to calculate a second intermediate coefficient and reconstructs a plurality of original pixel data. Claim 20 was abandoned upon payment of a registration fee. The method of claim 11, Step (b) is an image reconstruction method using the inverse discrete wavelet transform, characterized in that for performing the lifting-based inverse discrete wavelet transform. Claim 21 was abandoned upon payment of a registration fee. The method of claim 11, The step (c) is an image reconstruction method using inverse discrete wavelet transform, characterized in that for performing a lifting-based inverse discrete wavelet transform.

Images