High-dynamic infrared image compression method and device with variable scene adjustment

Abstract

The present invention provides a method and apparatus for high-dynamic infrared image compression with variable scene adjustment. The method includes the following steps: calculating the local variance matrix of the input infrared image data, and performing guided filtering on the input infrared image data according to the local variance matrix to obtain the base layer and detail layer of the image; performing global thresholding on the base layer to obtain the histogram matrix after global thresholding of the base layer; remapping the histogram matrix after global thresholding of the base layer according to the local variance matrix to obtain a generalized histogram matrix; remapping the generalized histogram matrix using Weber's law to obtain a new weight array, and obtaining the compressed image data of the base layer using the new weight array; and performing weighted combination of the compressed image data of the base layer and the detail layer data to obtain the compressed image data. The present invention adaptively allocates more mapping space for the gray levels in the regions with complex textures, and also has the characteristics of histogram equalization, avoiding over-suppression of large scenes.

Description

Technical Field

[0001] The present invention relates to the field of image processing, and in particular to a variable-scene adjustable high-dynamic infrared image compression method and apparatus. Background Art

[0002] With the gradual development of infrared imaging systems, in order to more accurately reflect the temperature differences in different regions of the scene to be measured, the number of bits of the original infrared image is usually 14 bits or 16 bits. However, ordinary display devices can usually only display 8-bit images. Therefore, it is necessary to compress the original infrared image, which is usually referred to as the High Dynamic Range Compression algorithm in this field. How to better retain and enhance the information of interest such as target details while compressing the number of bits of the image, and suppress the information that is not of interest and affects the visual effect of the image such as background noise is a problem that the field has been working on. Summary of the Invention

[0003] To solve at least some of the above problems existing in the prior art, the present invention provides a variable-scene adjustable high-dynamic infrared image compression method and apparatus.

[0004] The present invention is implemented as follows:

[0005] In a first aspect, the present invention provides a variable-scene adjustable high-dynamic infrared image compression method, including the following steps:

[0006] Calculate the local variance matrix of the input infrared image data, and perform guided filtering on the input infrared image data according to the local variance matrix to obtain the base layer and the detail layer of the image;

[0007] Perform global thresholding on the base layer to obtain the histogram matrix after global thresholding of the base layer;

[0008] According to the local variance matrix, remap the histogram matrix after global thresholding of the base layer to obtain a generalized histogram matrix;

[0009] Remap the generalized histogram matrix using Weber's law to obtain a new weight array, and use the new weight array to obtain the compressed image data of the base layer;

[0010] Perform weighted combination on the compressed image data of the base layer and the detail layer data to obtain the compressed image data.

[0011] Further, the formula for calculating the local variance matrix of the input infrared image data is as follows:

[0012] var=meanFilter(X 2 )-(meanFilter(X)) 2

[0013] In the formula, var is the local variance matrix, and meanFilter() is the mean filtering function with mirror padding at the boundary;

[0014] The guided filtering formula used to obtain the base layer and the detail layer of the image by guiding the filtering of the input infrared image data according to the local variance matrix is as follows:

[0015] dist = a.*src+(1 - a).*meanFilter(src)

[0016] a = var. / (var + esp)

[0017] In the formula, src is the input infrared image data, dist is the output filtered result matrix data, and esp is the filtering parameter.

[0018] Furthermore, the formula used to obtain the histogram matrix after global outlier removal of the base layer by performing global outlier removal on the base layer is as follows:

[0019]

[0020] In the formula, hist is the base layer histogram matrix, sHist is the histogram matrix after global outlier removal of the base layer, and llow is the outlier removal threshold.

[0021] Furthermore, the specific steps of obtaining the generalized histogram matrix by remapping the histogram matrix after global outlier removal of the base layer according to the local variance matrix include:

[0022] ① Map the local variance matrix so that the weight ratio of a single pixel after mapping is within the range of 0 to 1. The formula is as follows:

[0023]

[0024] ② Calculate the average weight ratio of each gray level in the histogram matrix after global outlier removal of the base layer. The formula is as follows:

[0025] ghVar(i) = mean(gVar*f)

[0026]

[0027] ③ Calculate the generalized histogram. The formula is as follows:

[0028] gHist(i) = min((sHist × ghVar)(i)+lMean,4 × lMean)

[0029] lMean = wSum / effHSize

[0030] wSum = min(sum(sHist × (1 - ghVar)), sum(sHist × ghVar) × (100 - gh) / 10)

[0031] In the above formulas, gh is an adjustable parameter with an adjustment range of 0 to 100, mean() is used to obtain the matrix mean, min(a, b) is used to take the smaller value of a and b, sum() is used to calculate the matrix sum, and effHSize is the number of non-zero values in the sHist array.

[0032] Further, the formula for obtaining the new weight array after remapping the generalized histogram matrix using Weber's law is as follows:

[0033]

[0034] In the formula, sW(i) is the weight of the i-th gray level in the new weight array, w(i) is the weight of the i-th gray level in the weight array, minI is the minimum gray level with non-zero gray levels, cot is the number of effective gray levels, and ace is an adjustable parameter with an adjustment range of -12 to 12.

[0035] Further, the input infrared image data is 14 bits, and the base layer compressed image data is 8 bits. Then the formula for obtaining the base layer compressed image data using the new weight array is as follows:

[0036] bu8Data(i, j) = sW(14Data(i, j)) / sumW * (maxV - minV) + minV

[0037] In the formula, bu8Data(i, j) is the 8-bit data at the (i, j) position of the obtained base layer, 14Data(i, j) is the 14-bit data at the (i, j) position of the original base layer, sum is the cumulative sum of the sW array, and maxV and minV are the maximum and minimum values mapped to 8-bit data.

[0038] Further, the formula for weighted merging of the base layer compressed image data and the detail layer data to obtain the compressed image data is as follows:

[0039] out8Data = bu8Data + sDetData × dde / 20

[0040] sDetData = detData. × var. / (var + d)

[0041] In the formula, out8Data is the compressed 8-bit image data, sDetDate is the weighted detail layer matrix, dde is an adjustable parameter with an adjustment value range of 0 to 100, detData is the original detail layer matrix, and d is the adaptive adjustment formula.

[0042] In a second aspect, the present invention provides a variable-scene adjustable high-dynamic infrared image compression device, including:

[0043] An original base layer and detail layer acquisition module, configured to calculate the local variance matrix of the input infrared image data, and perform guided filtering on the input infrared image data according to the local variance matrix to obtain the base layer and detail layer of the image;

[0044] A base layer global outlier module, configured to perform global outlier on the base layer to obtain the histogram matrix after global outlier of the base layer;

[0045] A generalized histogram matrix acquisition module, configured to remap the histogram matrix after global outlier of the base layer according to the local variance matrix to obtain a generalized histogram matrix;

[0046] A base layer compressed image data acquisition module, configured to remap the generalized histogram matrix using Weber's law to obtain a new weight array, and use the new weight array to obtain the base layer compressed image data;

[0047] A compressed image acquisition module, configured to perform weighted merging on the base layer compressed image data and the detail layer data to obtain the compressed image data.

[0048] In a third aspect, the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of any of the above methods are implemented.

[0049] In a fourth aspect, the present invention provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the steps of any of the above methods are implemented.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] The variable-scene adjustable high-dynamic infrared image compression method and device provided by the present invention can judge the texture complexity of the scenes where different gray values are located based on the local variance and histogram information of the image data, adaptively allocate more mapping space for the gray levels in the texture-complex regions, and also have the characteristics of histogram equalization. When high-temperature or low-temperature objects with a small image proportion enter, the mapping space allocated to them will be less than the linear mapping, avoiding over-suppression of large scenes. At the same time, three adjustable parameters with clear meanings are retained to adjust controversial and preference-based image metrics to meet the image requirements of more people in more scenes. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 FIG. is a flowchart of a variable-scene adjustable high-dynamic infrared image compression method provided by an embodiment of the present invention;

[0053] Figure 2 FIG. is a detailed flowchart of a variable-scene adjustable high-dynamic infrared image compression method provided by an embodiment of the present invention;

[0054] Figure 3 FIG. is a block diagram of a variable-scene adjustable high-dynamic infrared image compression device provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0055] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0056] As Figure 1 and Figure 2 shown, an embodiment of the present invention provides a variable-scene adjustable high-dynamic infrared image compression method. In this embodiment, taking the compression of 14-bit original infrared image data into 8-bit infrared image data as an example, the method includes the following steps:

[0057] S101. Calculate the local variance matrix of the input 14-bit infrared image data, and perform guided filtering on the input infrared image data according to the local variance matrix to obtain the base layer and detail layer of the image.

[0058] Among them, the formula for calculating the local variance matrix of the input 14-bit infrared image data is as follows:

[0059] var = meanFilter(X 2 ) - (meanFilter(X)) 2

[0060] In the formula, var is the local variance matrix, and meanFilter() is the mean filter function with mirror padding at the boundary; in this embodiment, the size of the mean filter template is 3*3. Substituting the input 14-bit infrared image data into X in the formula, the local variance matrix var can be obtained.

[0061] The guided filtering of the input infrared image data according to the local variance matrix to obtain the base layer and the detail layer of the image adopts the simplified guided filtering, and its guided filtering formula is:

[0062] dist = a.*src+(1-a).*meanFilter(src)

[0063] a = var. / (var+esp)

[0064] In the formula, src is the input infrared image data, dist is the output filtered result matrix data, and esp is the filtering parameter. The smaller its value, the better the edge-preserving effect and the worse the filtering effect; in this embodiment, when obtaining the base layer of the image, the guided filtering 1 parameter is taken as 900, and when obtaining the detail layer of the image, the guided filtering 2 parameter is taken as 20. In other embodiments, the selection of the guided filtering parameter is not limited to this and can be selected according to actual needs.

[0065] S102. Perform global dot throwing on the base layer to obtain the histogram matrix after global dot throwing of the base layer.

[0066] The specific formula used is as follows:

[0067]

[0068] In the formula, hist is the base layer histogram matrix, which can be obtained by processing the image base layer data obtained in step S101, sHist is the histogram matrix after global dot throwing of the base layer, llow is the dot throwing threshold, and its value calculation is represented by pseudocode:

[0069] a = 1;

[0070] count = 0;

[0071] while(count < dataSize*tRate)

[0072] {

[0073] if(hist <= a)

[0074] count++;

[0075] a++;

[0076] }

[0077] llow = a;

[0078] In the pseudocode, dataSize is the total number of image pixels, and tRata is the ratio of throwing points. The recommended value is 0.002.

[0079] S103. After globally throwing points on the histogram matrix of the base layer according to the local variance matrix, perform a new mapping to obtain a generalized histogram matrix; specifically including:

[0080] ① Map the local variance matrix so that the weight ratio of a single pixel after mapping is in the range of 0 to 1. The formula is as follows:

[0081]

[0082] ② Calculate the average weight ratio of each gray level in the histogram matrix after globally throwing points on the base layer. The formula is as follows:

[0083] ghVar(i) = mean(gVar * f)

[0084]

[0085] ③ Calculate the generalized histogram. The formula is as follows:

[0086] gHist(i) = min((sHist × ghVar)(i) + lMean, 4 × lMean)

[0087] lMean = wSum / effHSize

[0088] wSum = min(sum(sHist × (1 - ghVar)), sum(sHist × ghVar) × (100 - gh) / 10)

[0089] In the above formulas, gh is an adjustable parameter, and the adjustment range is 0 to 100. mean() is to obtain the matrix mean value, min(a, b) is to take the smaller value of a and b, sum() is to calculate the matrix sum, and effHSize is the number of non-zero values in the sHist array.

[0090] S104. After performing a new mapping on the generalized histogram matrix using Weber's law to obtain a new weight array, use the new weight array to obtain the compressed image data of the base layer.

[0091] Among them, the formula used to perform a new mapping on the generalized histogram matrix using Weber's law to obtain a new weight array is as follows:

[0092]

[0093] In the formula, sW(i) is the weight of the i-th gray level of the new weight array, w(i) is the weight of the i-th gray level of the weight array, minI is the minimum gray level with non-zero gray scale count, cot is the number of effective gray levels, ace is an adjustable parameter, and the adjustment value range is -12 to 12.

[0094] The formula for obtaining the base layer compressed image data using the new weight array is as follows:

[0095] bu8Data(i,j) = sW(14Data(i,j)) / sumW * (maxV - minV) + minV

[0096] In the formula, bu8Data(i,j) is the 8-bit data at the (i,j) position of the obtained base layer, 14Data(i,j) is the 14-bit data at the (i,j) position of the original base layer, sum is the cumulative sum of the sW array, maxV and minV are the maximum and minimum values mapped to 8-bit data, and their magnitudes are restricted by the mapping width. The calculation formula is:

[0097]

[0098]

[0099] In the formula, sizeW is the number of non-zero values in the sW array.

[0100] S105. Weightedly merge the base layer compressed image data and the detail layer data to obtain the compressed image data. The specific formula is as follows:

[0101] out8Data = bu8Data + sDetData × dde / 20

[0102] sDetData = detData. × var. / (var + d)

[0103] In the formula, out8Data is the compressed 8-bit image data, sDetDate is the weighted detail layer matrix, dde is an adjustable parameter, and the adjustment value range is 0 to 100. detData is the original detail layer matrix, that is, the image detail layer data obtained in step S101, and d is the adaptive adjustment formula.

[0104] In the above method, three adjustable parameters, dde, ace, and gh, are reserved to adjust controversial and preference-based image metrics to meet the image requirements of more people in more scenarios. Among them, dde is the image detail enhancement parameter. The larger the value, the greater the image detail enhancement, but the greater the noise will also be, and there may be over-enhancement at the edges, resulting in unnatural images. The larger the ace value, the greater the detail contrast in the high-temperature area of the image and the smaller the detail contrast in the low-temperature area, and the darker the image. Conversely, the smaller the detail contrast in the high-temperature area of the image and the greater the detail contrast in the low-temperature area, the brighter the image. The larger the gh value, the more the simple texture areas are suppressed and the more the complex texture areas are enhanced. When the gh value is 0, it is the histogram projection mapping.

[0105] The variable-scene adjustable high-dynamic infrared image compression method provided by the embodiment of the present invention can judge the texture complexity of the scene where different gray values are located based on the local variance and histogram information of the image data, adaptively allocate more mapping space for the gray levels in the complex texture areas, and also has the characteristics of histogram equalization. When high-temperature or low-temperature objects with a small image proportion enter, the allocated mapping space will be less than the linear mapping, avoiding over-suppression of large scenes. At the same time, three adjustable parameters with clear meanings are reserved to adjust controversial and preference-based image metrics to meet the image requirements of more people in more scenarios.

[0106] As Figure 3 shown, the embodiment of the present invention also provides a variable-scene adjustable high-dynamic infrared image compression device, including:

[0107] An original base layer and detail layer acquisition module 201, configured to calculate the local variance matrix of the input infrared image data, and perform guided filtering on the input infrared image data according to the local variance matrix to obtain the base layer and detail layer of the image;

[0108] A base layer global thresholding module 202, configured to perform global thresholding on the base layer to obtain the histogram matrix after global thresholding of the base layer;

[0109] A generalized histogram matrix acquisition module 203, configured to remap the histogram matrix after global thresholding of the base layer according to the local variance matrix to obtain a generalized histogram matrix;

[0110] A base layer compressed image data acquisition module 204, configured to remap the generalized histogram matrix using Weber's law to obtain a new weight array, and use the new weight array to obtain the base layer compressed image data;

[0111] A compressed image acquisition module 205, configured to perform weighted merging on the base layer compressed image data and the detail layer data to obtain the compressed image data.

[0112] An embodiment of the present invention further provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the above method embodiment are implemented.

[0113] An embodiment of the present invention further provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the steps of the above method embodiment are implemented.

[0114] Since the principles of the above device, electronic device, and computer-readable storage medium embodiments for solving the technology are similar to those of the above method embodiment, the implementation of the device, electronic device, and computer-readable storage medium can refer to the above method embodiment, and the repeated parts will not be described again.

[0115] The foregoing is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A high-dynamic infrared image compression method with variable scene adjustment, characterized in that It includes the following steps: Calculate the local variance matrix of the input infrared image data, and perform guided filtering on the input infrared image data according to the local variance matrix to obtain the base layer and the detail layer of the image; wherein, the formula for calculating the local variance matrix of the input infrared image data is as follows: var = meanFilter(X 2 ) - (meanFilter(X)) 2 In the formula, var is the local variance matrix, and meanFilter() is the mean filtering function with mirror padding at the boundary; The guided filtering formula used for performing guided filtering on the input infrared image data according to the local variance matrix to obtain the base layer and the detail layer of the image is: dist = a.*src+(1-a).*meanFilter(src) a = var. / (var+esp) In the formula, src is the input infrared image data, dist is the output filtered result matrix data, and esp is the filtering parameter; Perform global thresholding on the base layer to obtain the histogram matrix after global thresholding of the base layer, and the formula used is as follows: In the formula, hist is the base layer histogram matrix, sHist is the histogram matrix after global thresholding of the base layer, and llow is the threshold; According to the local variance matrix, perform re-mapping on the histogram matrix after global thresholding of the base layer to obtain the generalized histogram matrix, specifically including: ① Map the local variance matrix so that the weight ratio of a single pixel after mapping is within the range of 0 to 1, and the formula is as follows: ② Calculate the average weight ratio of each gray level in the histogram matrix after global thresholding of the base layer, and the formula is as follows: ghVar(i)=mean(gVar*f) ③ Calculate the generalized histogram, and the formula is as follows: gHist(i)=min((sHist×ghVar)(i)+lMean,4×lMean) lMean = wSum / effHSize wSum = min(sum(sHist×(1-ghVar)),sum(sHist×ghVar)×(100-gh) / 10) In the above formulas, gh is an adjustable parameter, and the adjustment range is 0 to 100, mean() is to obtain the matrix mean value, min(a,b) is to take the smaller value of a and b, sum() is to calculate the matrix sum, and effHSize is the number of non-zero values in the sHist array; Perform re-mapping on the generalized histogram matrix using Weber's law to obtain a new weight array, and use the new weight array to obtain the compressed image data of the base layer; wherein, the formula used for performing re-mapping on the generalized histogram matrix using Weber's law to obtain the new weight array is as follows: In the formula, sW(i) is the weight of the i-th gray level of the new weight array, w(i) is the weight of the i-th gray level of the weight array, minI is the smallest gray level with non-zero gray levels, cot is the number of effective gray levels, and ace is an adjustable parameter, and the adjustment range is -12 to 12; The input infrared image data is 14 bits, and the compressed image data of the base layer is 8 bits, then the formula for obtaining the compressed image data of the base layer using the new weight array is as follows: bu8Data(i,j) = sW(14Data(i,j)) / sumW * (maxV - minV) + minV In the formula, bu8Data(i,j) is the 8-bit data at the position (i,j) of the obtained base layer, 14Data(i,j) is the 14-bit data at the position (i,j) of the original base layer, sum is the cumulative sum of the sW array, and maxV and minV are the maximum and minimum values mapped to 8-bit data; The compressed image data of the base layer and the detail layer data are weighted and combined to obtain the compressed image data.

2. The variable-scene adjustable high-dynamic infrared image compression method according to claim 1, characterized in that, The formula for weighted combining the compressed image data of the base layer and the detail layer data to obtain the compressed image data is as follows: out8Data = bu8Data + sDetData × dde / 20 sDetData = detData. × var. / (var + d) In the formula, out8Data is the compressed 8-bit image data, sDetDate is the weighted detail layer matrix, dde is an adjustable parameter with an adjustment value range of 0 to 100, detData is the original detail layer matrix, and d is an adaptive adjustment formula.

3. A high-dynamic infrared image compression device with variable scene adjustment, characterized in that, Including: An original base layer and detail layer acquisition module, which is used to calculate the local variance matrix of the input infrared image data, and perform guided filtering on the input infrared image data according to the local variance matrix to obtain the base layer and detail layer of the image; among them, the formula for calculating the local variance matrix of the input infrared image data is as follows: var = meanFilter(X 2 ) - (meanFilter(X)) 2 In the formula, var is the local variance matrix, and meanFilter() is the mean filtering function with mirror filling at the boundary; The guided filtering formula used for performing guided filtering on the input infrared image data according to the local variance matrix to obtain the base layer and detail layer of the image is: dist = a.*src + (1 - a).*meanFilter(src) a = var. / (var + esp) In the formula, src is the input infrared image data, dist is the output filtered result matrix data, and esp is the filtering parameter; A base layer global dot throwing module, which is used to perform global dot throwing on the base layer to obtain the histogram matrix after global dot throwing of the base layer, and the formula used is as follows: In the formula, hist is the base layer histogram matrix, sHist is the histogram matrix after global dot throwing of the base layer, and llow is the dot throwing threshold; A generalized histogram matrix acquisition module, which is used to remap the histogram matrix after global dot throwing of the base layer according to the local variance matrix to obtain the generalized histogram matrix, specifically including: ① Map the local variance matrix so that the weight ratio of a single pixel after mapping is within the range of 0 to 1, and the formula is as follows: ② Calculate the average weight ratio of each gray level in the histogram matrix after global dot throwing of the base layer, and the formula is as follows: ghVar(i) = mean(gVar * f) ③ Calculate the generalized histogram, and the formula is as follows: gHist(i) = min((sHist × ghVar)(i) + lMean, 4 × lMean) lMean = wSum / effHSize wSum = min(sum(sHist × (1 - ghVar)), sum(sHist × ghVar) × (100 - gh) / 10). In the above formulas, gh is an adjustable parameter with an adjustment range of 0 to 100, mean() is used to calculate the matrix mean value, min(a, b) is used to take the smaller value of a and b, sum() is used to calculate the matrix sum, and effHSize is the number of non-zero values in the sHist array; The basic layer compressed image data acquisition module is used to obtain a new weight array after remapping the generalized histogram matrix using Weber's law, and obtain the basic layer compressed image data using the new weight array; among them, the formula used to obtain the new weight array after remapping the generalized histogram matrix using Weber's law is as follows: In the formula, sW(i) is the weight of the i-th gray level of the new weight array, w(i) is the weight of the i-th gray level of the weight array, minI is the smallest gray level with a non-zero number of gray levels, cot is the number of effective gray levels, and ace is an adjustable parameter with an adjustment range of -12 to 12; The input infrared image data is 14 bits, and the basic layer compressed image data is 8 bits. Then the formula for obtaining the basic layer compressed image data using the new weight array is as follows: bu8Data(i, j) = sW(14Data(i, j)) / sumW * (maxV - minV) + minV In the formula, bu8Data(i, j) is the 8-bit data at the position (i, j) of the obtained basic layer, 14Data(i, j) is the 14-bit data at the position (i, j) of the original basic layer, sum is the cumulative sum of the sW array, and maxV and minV are the maximum and minimum values mapped to 8-bit data; The compressed image acquisition module is used to perform weighted merging on the basic layer compressed image data and the detail layer data to obtain the compressed image data.

4. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method described in any one of claims 1-2.

5. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the method described in any one of claims 1-2.

Images