CN107481205A - A kind of Terahertz image fringes noise processing method and system - Google Patents

Abstract

The present invention relates to the field of terahertz detection array or infrared detection array imaging, and discloses a terahertz image fringe noise processing method and system, including the following content: Fourier transform is performed on the acquired terahertz image data to obtain and characterize the terahertz image data. A frequency-domain map of the frequency-domain characteristics of the Hertz image data; the first band-stop filtering is performed on the frequency-domain map to eliminate periodic noise in the frequency range where the streak noise is located; High-pass filtering is performed on the image to attenuate or suppress low-frequency components to highlight the remaining streak noise; a second band-stop filter is performed on the high-pass-filtered frequency-domain image to filter out the stripes that are not completely filtered out by the first band-stop filter The second band-stop filtering is performed on the noise; the frequency-domain image after the second band-stop filtering is converted into a time-domain image by inverse Fourier transform, and then the terahertz fringe noise can be effectively eliminated.

Description

A method and system for processing streak noise in terahertz images

technical field

The invention relates to the field of terahertz detection array or infrared detection array imaging, in particular to a method and system for processing fringe noise of a terahertz image.

Background technique

Terahertz radiation refers to electromagnetic waves with a frequency between 0.1THz and 10THz. Its wave band is between microwave and infrared, and it belongs to the category of far-infrared electromagnetic radiation. The terahertz spectrum of matter contains rich physical and chemical information. At the same time, due to the characteristics of terahertz radiation itself, it can become a complementary technology of Fourier transform infrared optical latent technology and X-ray technology in many aspects. Terahertz radiation has broad applications in many basic research fields, industrial applications and military applications. Expansion capacity. It also has great application potential in the fields of biology, medicine, microelectronics, agriculture and security inspection. In addition, compared with low-frequency electromagnetic waves, terahertz has a higher frequency and can be used as a communication carrier, which can carry more information per unit time. Terahertz radiation has good directivity and can be used for short-distance directional secure communication in the battlefield. Higher spatial resolution and longer depth of field can also be obtained by using terahertz imaging.

However, due to the characteristics of terahertz radiation itself and its long wavelength, interference fringe noise often occurs when using terahertz detectors for image acquisition. Researchers have done a lot of research on this, but most of them are still directly high-frequency Filtering cannot filter out interference noise in a targeted manner, so there are still many deficiencies. Moreover, due to the density and direction uncertainty of fringes during acquisition, it is still very difficult to remove interference fringes.

Therefore, the streak noise of the terahertz image cannot be effectively eliminated in the prior art.

Contents of the invention

In order to solve the technical problem in the prior art that the fringe noise of a terahertz image cannot be effectively eliminated, the present invention further provides a method and system for processing the fringe noise of a terahertz image.

In order to solve the above-mentioned technical problems, a technical solution adopted by the present invention is: a method for processing terahertz image streak noise, including the following content:

Performing Fourier transform on the acquired terahertz image data to obtain a frequency domain map representing the frequency domain characteristics of the terahertz image data;

Carrying out the first band-stop filtering on the frequency-domain image to eliminate the periodic noise in the frequency range where the streak noise is located;

Perform high-pass filtering on the frequency domain image after the first band-stop filtering to attenuate or suppress low-frequency components and highlight the remaining streak noise;

Performing a second band-stop filter on the high-pass filtered frequency-domain image is used to perform a second band-stop filter on the streak noise that was not completely filtered out by the first band-stop filter;

The frequency-domain image after the second band-stop filtering is transformed into a time-domain image by inverse Fourier transform.

On the other hand, a terahertz image streak noise processing system is also provided, including:

A Fourier transform module, configured to perform Fourier transform on the terahertz image data to obtain a frequency-domain map representing the frequency-domain features of the terahertz image data;

The first band-stop filter is used to perform a first band-stop filter on the frequency domain image to eliminate periodic noise in the frequency range where the streak noise is located;

A frequency-domain high-pass filter is used to perform high-pass filtering on the frequency-domain image after the first band-stop filter, to attenuate or suppress low-frequency components, and to highlight the remaining streak noise;

The second band-stop filter is used to perform a second band-stop filter on the high-pass filtered frequency domain image, and perform a second band-stop filter on the first band-stop filter that does not filter out complete streak noise;

The inverse Fourier transform module is used to transform the frequency-domain image after the second band-stop filtering into a time-domain image by using an inverse Fourier transform.

The beneficial effect of the present invention is: be different from the situation of prior art:

In this terahertz image streak noise processing method, Fourier transform is performed on the acquired terahertz image data, and then part of the streak noise is removed through the first band-stop filter, and then, after high-pass filtering, the second band-stop filter Remove the remaining fringe noise, and finally convert it into a time-domain image through inverse Fourier transform to obtain the final noise-filtered image. The image highlights the target object, and the filter can be set through the frequency-domain image, which is simple and easy to observe, and solves the problem. The problem of the generality of the algorithm, and finally through image enhancement, the target object is more obvious, so that the processed image not only eliminates the stripe noise, but also plays the role of image enhancement.

Description of drawings

FIG. 1 is a schematic flow chart of the steps of a method for processing streak noise in a terahertz image in an embodiment of the present invention;

Fig. 2a-Fig. 2f are the effect diagrams of the fringe noise processing process of the terahertz image in the embodiment of the present invention;

Fig. 3 is a block diagram of a terahertz image streak noise processing system in an embodiment of the present invention.

detailed description

In order to solve the technical problem that the fringe noise of the terahertz image cannot be effectively eliminated in the prior art, the present invention further provides a method and system for processing the fringe noise of the terahertz image, which can effectively eliminate the fringe noise of the terahertz image .

In order to better understand the technical solution of the present invention, the technical solution of the present invention will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

A method for processing streak noise in a terahertz image provided by an embodiment of the present invention, as shown in FIG. 1 , includes: S101, performing Fourier transform on the acquired terahertz image data to obtain a frequency domain characterizing the terahertz image data The frequency-domain image of the feature; S102, performing the first band-rejection filter on the frequency-domain image to eliminate the periodic noise in the frequency range where the streak noise is located; S103, performing high-pass on the frequency-domain image after the first band-rejection filter Filtering, used to attenuate or suppress low-frequency components, highlighting the remaining streak noise; S104, performing a second band-stop filter on the high-pass filtered frequency domain image, used to filter out the streak noise that was not completely filtered out by the first band-stop filter Perform a second band-rejection filtering; S105, convert the frequency-domain image after the second band-rejection filtering into a time-domain image by inverse Fourier transform.

In a specific implementation, the acquired terahertz image data can be the image data that has been acquired by the terahertz imaging system and stored in the memory, read from the memory, or it can be the image currently acquired by the terahertz imaging system Data, the specific original image is shown in Figure 2a.

Specifically, Fourier transform is performed on the acquired terahertz image data to obtain a frequency domain map representing the frequency domain characteristics of the terahertz image data, which specifically includes: using Fourier transform to shift the frequency spectrum of the acquired terahertz image data frequency to the origin, so that the frequency distribution of the terahertz image is symmetrically distributed with the origin as the center; then, the frequency distribution is obtained from the Fourier-transformed terahertz image, in addition to the bright spot at the center of the circle, there is also a set of bright spots with symmetrical distribution, The set of bright spots is generated by interference noise and a periodic interference signal.

The frequency distribution of the image can be clearly seen when the frequency spectrum is shifted to the center of the circle above, and the periodic interference signal can be separated. It can be seen from the frequency shift to the origin of the spectrum diagram that there is a certain point as the center in addition to the center of the circle. Symmetrically distributed bright spot set, this bright spot set is generated by interference noise, at this time, it is very intuitive to eliminate the interference by placing a band-stop filter at this position. A specific frequency domain diagram obtained after Fourier transform is shown in FIG. 2b.

Since the frequency domain map obtained in S101 can see the frequency domain range where the interference fringe noise is located, therefore, in S102, a suitable first band-stop filter can be constructed, and the formula corresponding to the constructed first band-stop filter Multiply with the formula corresponding to the Fourier transform of the terahertz image data to filter out most of the fringe noise components in the frequency domain.

The first band-stop filter is used to suppress the frequency of a ring area at a certain distance from the center of the frequency domain, and can be used to eliminate periodic noise in a certain frequency range. For the first band-stop filter that performs the first band-stop filtering selection, the specific steps are as follows:

The formula from this bandstop filter is:

Among them, D ₀ is the distance between the frequency point to be blocked and the frequency center, W is the bandwidth of the band-stop filter, and for an image of size M*N, the distance between the frequency point (u, v) and the frequency domain center is D ₀ (u,v), whose expression is H ₀ (u, v) is the required band-stop filter formula, when its value is 1, the band under this frequency domain is completely passed, when its value is 0, the band under this frequency domain is completely filtered . According to the range of the frequency domain, the distance D ₀ between the frequency point to be blocked and the center of the frequency domain and the bandwidth W of the band-stop filter are obtained. Thus, parameters of the first bandpass filter are obtained.

Next, in S103 , high-pass filtering is performed on the frequency-domain image that has undergone the first band-rejection filtering, so as to attenuate or suppress low-frequency components and highlight remaining streak noise. Specifically, the Butterworth filter is used to perform second-order high-pass filtering on the frequency-domain image that has undergone the first band-rejection filter, which is used to attenuate or suppress low-frequency components and highlight the remaining streak noise. Specifically, the Butterworth The filter is a type of filter in the Fourier frequency domain. The gradient of the truncated part of the transfer function of the Butterworth filter can be controlled by the exponent n, and the truncated part of the low-order Butterworth filter will not be very Steep, the ringing effect can be mitigated or avoided. The characteristic of the Butterworth filter is that the frequency response curve in the same frequency band is as flat as possible without fluctuations, and it gradually decreases to zero in the stop frequency band. On the Bode plot of the logarithmic diagonal frequency of the amplitude, starting from a certain boundary angular frequency, the amplitude gradually decreases with the increase of the angular frequency, tending to negative infinity, and the attenuation rate of the second-order Butterworth filter per time Frequency 12dB.

Specifically, the transfer function of the Butterworth high-pass filter is

Among them, D ₁ is the cut-off frequency of the Butterworth filter, n is the order of the Butterworth filter, which is used to control the steepness of the Butterworth filter. For an image with a size of M*N, the frequency points (u , v) is D ₁ (u,v) from the center of the frequency domain, and its expression is

The frequency domain diagram after the above-mentioned Butterworth filter is shown in Figure 2c. It can be seen that some of the interference fringes have not been filtered out, and the frequency domain range of these interference fringes can be found through the frequency domain diagram at this time. , which is convenient for re-filtering.

Therefore, in S104, a second band-stop filter is performed on the high-pass filtered frequency-domain image, which is used to perform a second band-stop filter on the streak noise that was not completely filtered out by the first band-stop filter, and the obtained As shown, the second band-stop filter used in the second band-stop filtering also needs to be constructed at this time. Since the interference fringes are closely arranged during the first band-stop filtering process, the photoelectric position on the frequency domain map is far from the center origin Closer and larger in diameter, therefore, the distance between the frequency point to be blocked and the frequency center selected in the first band-stop filter is small, and the bandwidth W value is large; in the second band-stop filtering process, the interference fringe arrangement Loose, the position of the photoelectricity on the frequency domain diagram is far away from the center origin, and the diameter is small, so the distance between the frequency point to be blocked and the frequency center selected by the second band-stop filter is relatively large, and the bandwidth W value is small. Specifically, the distance between the frequency point that needs to be blocked and the frequency center selected by the second band-stop filter is greater than the distance between the frequency point that needs to be blocked and the frequency center selected by the first band-stop filter, and the second band-stop filter The bandwidth of is smaller than the bandwidth of the first band-stop filter.

Finally, S105 is executed to convert the frequency-domain image after the second band-stop filtering into a time-domain image by inverse Fourier transform. Specifically shown in Figure 2e.

It can be seen from Figure 2e that the contrast of the image is relatively low. In order to emphasize the overall or local characteristics of the image, the original unclear image becomes clear or some interesting features are emphasized, and the difference between different object features in the image is enlarged. Suppress uninteresting features to improve image quality, enrich information, enhance image interpretation and recognition effects, and meet the needs of some special analysis. Therefore, after inverse Fourier transform, adjust the grayscale of the time domain image range, the graph shown in Figure 2f is obtained.

Specifically, linear grayscale stretching is performed on the time domain image within the grayscale range of 0-255. Specifically, when operating on the pixels in the image, the formula is described as follows:

g(x,y)=f(x,y)*h(x,y), where f(x,y) is the original image; h(x,y) is the space conversion function; g(x,y) Indicates the processed image. to get the final result graph.

The image after the above processing not only eliminates the stripe noise, but also plays the role of image enhancement.

Based on the same inventive concept, an embodiment of the present invention also provides a terahertz image streak noise processing system, as shown in FIG. device 303, a second band-rejection filter 304, and an inverse Fourier transform module 305, wherein the Fourier transform module 301 is used to perform Fourier transform on the acquired terahertz image data to obtain a representation of the terahertz image data The frequency-domain graph of frequency-domain features; the first band-stop filter 302, used to carry out band-stop filtering for the first time to the frequency-domain graph, to eliminate the periodic noise in the frequency range where the streak noise is located, and the frequency-domain high-pass filter 303, used for Perform high-pass filtering on the frequency-domain image after the first band-stop filtering, for attenuating or suppressing low-frequency components, highlighting the remaining streak noise, and the second band-stop filter 304, for performing the first high-pass filtering on the frequency-domain image after high-pass filtering The second band-stop filtering is to perform the second band-stop filtering on the band-stop filtering that does not filter out the complete streak noise in the first band-stop filtering, and the Fourier inverse transform module 305 is used to convert the frequency domain through the second band-stop filtering The graph is transformed into a time-domain graph using an inverse Fourier transform.

In this specific embodiment, the terahertz image streak noise processing system further includes an image enhancement module 306, configured to adjust the grayscale range of the time-domain image obtained by the inverse Fourier transform module, so as to obtain an enhanced image.

The above is only an embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process conversion made by using the description of the present invention and the contents of the accompanying drawings, or directly or indirectly used in other related technical fields , are all included in the scope of patent protection of the present invention in the same way.

Claims (9)

1. a kind of Terahertz image fringes noise processing method, it is characterised in that including following content：

Fourier transformation is carried out to the Terahertz view data of acquisition, obtains the frequency domain character for characterizing the Terahertz view data Frequency domain figure；

First time bandreject filtering is carried out to the frequency domain figure, for eliminating the periodic noise of the frequency range where fringes noise；

It is prominent remaining for decaying or suppressing low frequency component to carrying out high-pass filtering by the frequency domain figure of first time bandreject filtering Fringes noise；

Second of bandreject filtering is carried out to the frequency domain figure Jing Guo high-pass filtering, for not filtered out completely to first time bandreject filtering Fringes noise carries out second of bandreject filtering；

Inverse Fourier transform will be used by the frequency domain figure of second of bandreject filtering, be converted to time-domain diagram.

2. Terahertz image fringes noise processing method according to claim 1, it is characterised in that passing through by described The frequency domain figure of secondary bandreject filtering uses inverse Fourier transform, after being converted to time-domain diagram, in addition to：

Tonal range is adjusted to the time-domain diagram.

3. Terahertz image fringes noise processing method according to claim 1, it is characterised in that described pair obtains too Hertz view data carries out Fourier transformation, obtains the frequency domain figure for the frequency domain character for characterizing the Terahertz view data, specifically Including：

The frequency spectrum shift frequency of the Terahertz view data obtained using Fourier transform pairs is to origin so that the Terahertz image Frequency distribution is symmetrical using origin as the center of circle；

Frequency distribution is obtained from the Terahertz image Jing Guo Fourier transformation, in addition to the center of circle bright spot, also there is symmetrical point The bright spot set of cloth, the bright spot collection is combined into caused by interfering noise, and the interference signal of periodic regularity.

4. Terahertz image fringes noise processing method according to claim 1, it is characterised in that enter to the frequency domain figure Row first time bandreject filtering, for eliminating the periodic noise of the frequency range where fringes noise, specifically include：

Pair so frequency domain figure is found out on the centrosymmetric bright spot of frequency domain, it is determined that what is needed to use is corresponding with first time bandreject filtering The first bandstop filter；

First bandstop filter is used for the frequency for suppressing a circle ring area apart from frequency domain center pre-determined distance, wherein institute The formula for stating bandstop filter is：

<mrow> <msub> <mi>H</mi> <mn>0</mn> </msub> <mrow> <mo>(</mo> <mi>u</mi> <mo>,</mo> <mi>v</mi> <mo>)</mo> </mrow> <mo>=</mo> <mfenced open = "{" close = ""> <mtable> <mtr> <mtd> <mrow> <mn>1</mn> <mo>,</mo> <msub> <mi>D</mi> <mn>0</mn> </msub> <mrow> <mo>(</mo> <mi>u</mi> <mo>,</mo> <mi>v</mi> <mo>)</mo> </mrow> <mo>&lt;</mo> <msub> <mi>D</mi> <mn>0</mn> </msub> <mo>-</mo> <mfrac> <mi>W</mi> <mn>2</mn> </mfrac> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <mn>0</mn> <mo>,</mo> <msub> <mi>D</mi> <mn>0</mn> </msub> <mo>-</mo> <mfrac> <mi>W</mi> <mn>2</mn> </mfrac> <mo>&amp;le;</mo> <msub> <mi>D</mi> <mn>0</mn> </msub> <mrow> <mo>(</mo> <mi>u</mi> <mo>,</mo> <mi>v</mi> <mo>)</mo> </mrow> <mo>&amp;le;</mo> <msub> <mi>D</mi> <mn>0</mn> </msub> <mo>+</mo> <mfrac> <mi>W</mi> <mn>2</mn> </mfrac> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <mn>1</mn> <mo>,</mo> <msub> <mi>D</mi> <mn>0</mn> </msub> <mrow> <mo>(</mo> <mi>u</mi> <mo>,</mo> <mi>v</mi> <mo>)</mo> </mrow> <mo>&gt;</mo> <msub> <mi>D</mi> <mn>0</mn> </msub> <mo>+</mo> <mfrac> <mi>W</mi> <mn>2</mn> </mfrac> </mrow> </mtd> </mtr> </mtable> </mfenced> </mrow>

Wherein, D₀To need the distance of the Frequency point and center frequency prevented, W is the bandwidth of bandstop filter, is M* for size N image, Frequency point (u, v) and the distance at frequency domain center are D₀(u, v), its expression formula are

Formula corresponding to first bandstop filter is multiplied with corresponding formula after Terahertz view data Fourier transformation, Filter out the periodic noise in the frequency domain where fringes noise.

5. Terahertz image fringes noise processing method according to claim 1, it is characterised in that to passing through first time band The frequency domain figure of resistance filtering carries out high-pass filtering, for decaying or suppressing low frequency component, prominent remaining fringes noise, is specially：

Second order high-pass filtering processing is carried out using Butterworth filter to the frequency domain figure by first time bandreject filtering, for declining Subtract or suppress low frequency component, prominent remaining fringes noise, the transmission function formula of the Butterworth filter is specially：

<mrow> <msub> <mi>H</mi> <mn>1</mn> </msub> <mrow> <mo>(</mo> <mi>u</mi> <mo>,</mo> <mi>v</mi> <mo>)</mo> </mrow> <mo>=</mo> <mfrac> <mn>1</mn> <mrow> <mn>1</mn> <mo>+</mo> <msup> <mrow> <mo>&amp;lsqb;</mo> <msub> <mi>D</mi> <mn>1</mn> </msub> <mo>/</mo> <msub> <mi>D</mi> <mn>1</mn> </msub> <mrow> <mo>(</mo> <mi>u</mi> <mo>,</mo> <mi>v</mi> <mo>)</mo> </mrow> <mo>&amp;rsqb;</mo> </mrow> <mn>2</mn> </msup> </mrow> </mfrac> </mrow>

Wherein, D₁For the cut-off frequency of Butterworth filter, n is the exponent number of Butterworth filter, for controlling the Bart The steep of Butterworth wave filter, for the image that size is M*N, Frequency point (u, v) and the distance at frequency domain center are D₁(u, V), its expression formula is

<mrow> <msub> <mi>D</mi> <mn>1</mn> </msub> <mrow> <mo>(</mo> <mi>u</mi> <mo>,</mo> <mi>v</mi> <mo>)</mo> </mrow> <mo>=</mo> <msup> <mrow> <mo>&amp;lsqb;</mo> <msup> <mrow> <mo>(</mo> <mi>u</mi> <mo>-</mo> <mfrac> <mi>M</mi> <mn>2</mn> </mfrac> <mo>)</mo> </mrow> <mn>2</mn> </msup> <mo>+</mo> <msup> <mrow> <mo>(</mo> <mi>v</mi> <mo>-</mo> <mi>N</mi> <mo>/</mo> <mn>2</mn> <mo>)</mo> </mrow> <mn>2</mn> </msup> <mo>&amp;rsqb;</mo> </mrow> <mrow> <mn>1</mn> <mo>/</mo> <mn>2</mn> </mrow> </msup> <mo>.</mo> </mrow>

6. Terahertz image fringes noise processing method according to claim 4, it is characterised in that the second bandstop filter The Frequency point that prevents of needs and the distance of center frequency be more than Frequency point and the frequency that the needs of the first bandstop filter prevent The distance at center, the bandwidth of second bandstop filter are less than the bandwidth of the first bandstop filter, second bandreject filtering Device is that second of bandreject filtering uses, and the first bandstop filter is what first time bandreject filtering used.

7. Terahertz image fringes noise processing method according to claim 2, it is characterised in that adjusted to the time-domain diagram Whole tonal range, it is specially：

Linear gradation stretching is done in 0-255 tonal range to the time-domain diagram.

A kind of 8. Terahertz image fringes noise processing system, it is characterised in that including：

Fourier transformation module, for carrying out Fourier transformation to the Terahertz view data of acquisition, obtain and characterize the terahertz The hereby frequency domain figure of the frequency domain character of view data；

First bandstop filter, for carrying out first time bandreject filtering to the frequency domain figure, eliminate the frequency where fringes noise The periodic noise of scope；

Frequency domain high-pass filter, for carrying out high-pass filtering to the frequency domain figure for passing through first time bandreject filtering, for decaying or pressing down Low frequency component processed, prominent remaining fringes noise；

Second bandstop filter, for carrying out second of bandreject filtering to the frequency domain figure Jing Guo high-pass filtering, band for the first time is hindered Filtering does not filter out second of bandreject filtering of carry out of complete fringes noise；

Inverse Fourier transform module, for inverse Fourier transform will to be used by the frequency domain figure of second of bandreject filtering, be converted to Time-domain diagram.

9. Terahertz image fringes noise processing system according to claim 8, it is characterised in that also include：

Image enhancement module, for adjusting tonal range to the time-domain diagram.