Method for improving detail quality of high-dynamic infrared digital image based on weighted gamma correction

Abstract

The invention relates to a weighted gamma correction-based high-dynamic infrared digital image detail quality improvement method, which belongs to the technical field of image processing, solves the technical problem of excessive enhancement or detail loss in infrared image processing, and comprises the following steps of: reading an original high-bit deep infrared image I, performing histogram analysis on the original high-bit deep infrared image I, and obtaining an original high-bit deep infrared image II; weight distribution is carried out on the probability distribution function pdf (1), accumulation calculation is carried out on the weighted probability distribution function pdfw (1), weighted gamma correction is carried out on the original high-order deep infrared image I, filtering processing is carried out on the original high-order deep infrared image I, difference calculation is carried out on the filtered high-order deep infrared image I ', and the high-order deep infrared image I' is obtained. And carrying out fusion processing on the infrared image T'after gray mapping and the detail texture information Id 'to obtain an enhanced displayable infrared image Ie. The method is used for high-bit deep infrared image processing.

Description

Technical Field

[0001] The present invention belongs to the technical field of image processing, and in particular relates to a method for improving detail quality of high dynamic infrared digital images based on weighted gamma correction. Background Art

[0002] Compared with visible light detectors, infrared thermal imaging detectors have the characteristics of good anti-interference, strong target recognition ability, and all-weather operation. Therefore, they have very important application value in aerospace, sea surface detection, military monitoring and other fields. However, infrared band images usually have various noises due to the unsatisfactory optical performance of infrared detectors, randomness in the manufacturing process, and readout circuits, resulting in low image contrast and poor visual quality. In addition, infrared imaging often has high dynamic range information, and the varying degrees of image degradation that are prevalent in complex scenes can easily blur the edge texture information of the image, making the observed target submerged in the complex background. These defects seriously affect the infrared detection system's perception of the environment.

[0003] At the same time, infrared band images need to be grayscale mapped to 8bit before they can be used due to their high dynamic range grayscale, but random noise, image bad pixels and other phenomena will cause image mapping instability and unclear stretching during grayscale mapping. Therefore, designing an image detail quality improvement method to retain the details of high dynamic range images, overcome the difficulty of finding low-contrast targets in high dynamic range scenes, and perform efficient and stable image grayscale stretching display is extremely important for improving the application value of infrared images.

[0004] In response to the above problems, some researchers have proposed relevant infrared image detail enhancement methods. The Chinese invention patent with publication number CN113096053A discloses a high-dynamic infrared image detail enhancement method based on multi-scale guided filtering. The method adopts a layered idea, performs weighted fusion on the image detail layers of different scales obtained by multi-scale guided filtering decomposition, performs nonlinear compression mapping on the multi-scale fused detail layer and the corresponding basic layer, and normalizes the display to obtain the enhanced image; the Chinese invention patent with publication number CN110047055B discloses an infrared image detail enhancement and denoising method. The method performs layered processing on the infrared image, performs improved histogram equalization processing on the background layer, performs Gaussian filtering, pulse point determination, and Gamma transformation on the detail layer, and finally obtains the output image through weighted fusion.

[0005] In summary, the existing commonly used technologies and methods still have shortcomings in the following aspects:

[0006] 1. The image enhancement effect of existing infrared image detail enhancement methods is often not guaranteed, and the effect is not controllable in different application scenarios. There is a phenomenon of over-enhancement or detail loss, which cannot well meet the requirements of actual applications for image detail quality;

[0007] 2. The existing infrared image detail enhancement methods have high algorithm complexity and large requirements on computing resources. In practical applications, they often cannot meet the real-time requirements in high frame rate scenarios on embedded platforms. Summary of the invention

[0008] In order to overcome the shortcomings of over-enhancement or detail loss and complex algorithms in infrared image processing, the present invention proposes a method for improving the detail quality of high dynamic infrared digital images based on weighted gamma correction.

[0009] The technical solution adopted by the present invention to solve the technical problem is:

[0010] A method for improving the detail quality of high dynamic infrared digital images based on weighted gamma correction comprises the following steps:

[0011] Step 1: read the original high-bit-depth infrared image I, the bit depth of the image may be 10 / 12 / 14 / 16 bits, and the width and height of the image I are W and H respectively;

[0012] Step 2: Perform a histogram analysis on the read original high-bit deep infrared image I to obtain the probability distribution function pdf(l);

[0013] Step 3: Perform weight distribution on the probability distribution function pdf(l) obtained by statistics to obtain the weighted probability distribution function pdf w (l);

[0014] Step 4: The weighted probability distribution function pdf obtained in step 3 w (l) performing cumulative calculation to obtain the cumulative distribution function cdf(l); performing weighted gamma correction on the original high-bit deep infrared image I by using the cumulative distribution function cdf(l) to obtain a high-bit deep infrared image T after contrast enhancement;

[0015] Step 5: filtering the original high-bit deep infrared image I to obtain a filtered high-bit deep infrared image I′;

[0016] Step 6: Perform differential calculation on the filtered high-bit deep infrared image I′ to obtain the detail texture information I of the original high-bit deep infrared image I d ;

[0017] Step 7: Detail texture information I of the infrared image obtained in step 6 d Do truncation processing;

[0018] Step 8: The contrast-enhanced high-bit deep infrared image T and the infrared image detail texture information I obtained in step 5 and step 7 are respectively d Perform the weighted probability distribution function pdf obtained in step 3 w (l) grayscale mapping, respectively obtain the displayable infrared image T′ and the detail texture information I d ′;

[0019] Step 9: The infrared image T′ and the detail texture information I obtained after grayscale mapping in step 8 are d ′ Perform fusion processing to obtain the enhanced displayable infrared image I e ;

[0020] Further, the probability distribution function pdf(l) obtained in step 2 is as follows:

[0021]

[0022] Where n l represents the number of pixels with a grayscale value of 1 in the original high-bit deep infrared image I;

[0023] Furthermore, the weighted probability distribution function pdf obtained in step 3 is w (l) as follows;

[0024]

[0025] In the formula, pdf max 、pdf min They represent the maximum and minimum values ​​of the probability distribution function pdf(l) respectively; α represents an adjustable power coefficient, α>0, and the larger the α value, the more obvious the enhancement effect; β represents a linear coefficient;

[0026] Furthermore, the weighted probability distribution function pdf obtained in step 3 is w The linear coefficient β in (l) is as follows:

[0027]

[0028] Where, X M , X G , X max , X min are respectively represented by the mean, median, maximum and minimum values ​​of the grayscale values ​​of the original high-bit deep infrared image I pixel, wherein the median X G Represented as X G =(X max +X min ) / 2;

[0029] Further, the cumulative distribution function cdf(l) obtained in step 4 is as follows:

[0030]

[0031] Where, Σpdf w Expressed as the weighted probability distribution function pdf w The cumulative sum of (l); l max Expressed as the weighted probability distribution function pdf w (l) The maximum gray value l in (l);

[0032] Furthermore, the contrast-enhanced high-bit deep infrared image T obtained in step 4 is as follows;

[0033]

[0034] Where, T n (l) represents the pixel gray value after the weighted gamma correction corresponding to the pixel gray value l;

[0035] Furthermore, the filtered high-bit deep infrared image I′ obtained in step 5 is obtained by filtering the original high-bit deep infrared image I, and the calculation method is as follows;

[0036] I′(x,y)=blur(I(x,y))

[0037] In the formula, x and y represent the position coordinates of the pixel point;

[0038] Furthermore, the filtering method in step 5 includes but is not limited to Gaussian filtering, median filtering, edge filtering, guided filtering and other filtering methods. Considering the complexity of the algorithm, Gaussian filtering can be selected to maintain a good filtering effect while taking up less computing time;

[0039] Furthermore, the detail texture information I obtained in step 6 d It is obtained by performing a differential calculation between the original high-bit deep infrared image I and the filtered high-bit deep infrared image I′, and the calculation method is as follows;

[0040] I d (x,y)=I(x,y)-I′(x,y)

[0041] Furthermore, step 7 is to process the infrared image detail texture information I d The truncation process is performed to purify the detailed texture information of the infrared image, and the calculation method is as follows;

[0042]

[0043] In the formula, ε is expressed as a constant, and the size of ε determines the strength of background purification;

[0044] Furthermore, step 7 can be omitted, and the infrared image detail texture information I is not d When doing processing, you can sacrifice a certain purification effect in exchange for lower algorithm complexity and reduce algorithm processing time;

[0045] Furthermore, step 8 is based on the weighted probability distribution function pdf obtained in step 3. w The grayscale mapping of (l) is calculated as follows;

[0046]

[0047]

[0048] Where, T max 、T min Respectively expressed as the weighted probability distribution function pdf w (l) in pdf w (l)≥δ,l=0,1...l max 、pdf w (l)≥δ,l=l max ,l max The l value obtained when -1,...0, where the parameter δ is an adjustable parameter; I dmax ,I dmin Same reason.

[0049] Furthermore, the grayscale mapping parameter δ in step 8 is an adjustable parameter used to control the brightness of the image after grayscale mapping. The parameter can be an adjustable value t, δ=t, or an adjustable thousandth of the total number of pixels t‰, δ=t*W*H‰, for T max and T min The adjustable parameter δ can be taken as different parameter values ​​by adjusting the numerical value t or the thousandth ratio t‰;

[0050] Furthermore, the infrared image T′ and the detail texture information I after grayscale mapping in step 9 d ′ are all 8-bit images, which can be directly fused and displayed. The fusion method is weighted processing, and the calculation method is as follows;

[0051] I e (x,y)=λT′(x,y)+γI d ′(x,y)

[0052] Where, parameters λ and γ are adjustable coefficients, λ∈[0,1], γ∈[0,1];

[0053] Furthermore, if the original high-bit deep infrared image I read in step 1 is 8 bits, then the grayscale mapping in step 8 is only for the infrared image detail texture information I d Execute, and the remaining steps are executed normally to improve the image quality.

[0054] The beneficial effects of the present invention are:

[0055] A method for improving the detail quality of high-dynamic infrared digital images based on weighted gamma correction has low algorithm time complexity. It can significantly improve the quality of high-bit depth infrared images and be effectively applied to real-time processing based on embedded platforms in high frame rate scenarios.

[0056] A method for improving the detail quality of high dynamic infrared digital images based on weighted gamma correction maintains the median grayscale of the original image unchanged after enhancement processing, solves the mean shift problem in which there is a significant difference between the average brightness of the input image and the average brightness of the output image, and avoids the mapping instability and unclear mapping problems caused by image noise and bad pixels in conventional mapping methods.

[0057] A method for improving the detail quality of high dynamic infrared digital images based on weighted gamma correction is proposed. The method reserves an adjustable power coefficient α for image contrast enhancement and a grayscale stretching brightness coefficient δ. In practical applications, the method can be adjusted according to different target scenes and application requirements, and has good adaptability and strong robustness. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 This is a flow chart of a method for improving detail quality of high dynamic infrared digital images based on weighted gamma correction according to Embodiment 1 of the present invention;

[0059] Figure 2 This is a diagram showing the effect of using different contrast enhancement adjustable power coefficients α in Example 1 of the present invention;

[0060] Figure 3 This is a diagram showing the effect of using different grayscale stretching brightness coefficients δ in Example 1 of the present invention;

[0061] Figure 4 The figure is a comparison diagram of the processing effect diagram (right) and the original image (left) of 4 different test images according to Example 1 of the present invention;

[0062] Figure 5 This is a comparison diagram of the effect diagram (bottom) and the original image (top) of the digital image detail quality improvement method according to Example 2 of the present invention applied to an 8-bit image with an adjustable power coefficient α of contrast enhancement of 0.5. DETAILED DESCRIPTION

[0063] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0064] Example 1

[0065] A method for improving the detail quality of high dynamic infrared digital images based on weighted gamma correction is proposed. The weighted gamma correction is used to correct the contrast of the original image. The original image is filtered and differentially operated to obtain the image texture details. The two are gray-scale mapped based on histogram analysis. Finally, the enhanced displayable infrared image is obtained by image fusion. The specific operation steps are as follows.

[0066] Step 1: read the original high-bit-depth infrared image I, the bit depth of the image may be 10 / 12 / 14 / 16 bits, and the width and height of the image I are W and H respectively;

[0067] Step 2: Perform a histogram analysis on the read original high-bit deep infrared image I to obtain the probability distribution function pdf(l), and the calculation formula is as follows;

[0068]

[0069] Where n l represents the number of pixels with a grayscale value of 1 in the original high-bit deep infrared image I;

[0070] Step 3: Perform weight distribution on the probability distribution function pdf(l) obtained by statistics to obtain the weighted probability distribution function pdf w (l), the calculation formula is as follows;

[0071]

[0072]

[0073] In the formula, pdf max 、pdf min They represent the maximum and minimum values ​​of the probability distribution function pdf(l) respectively; α represents the adjustable power coefficient, α>0, the larger the α value, the more obvious the enhancement effect; β represents the linear coefficient, X M , X G , X max , X min are respectively represented by the mean, median, maximum and minimum values ​​of the grayscale values ​​of the original high-bit deep infrared image I pixel, wherein the median X G Represented as X G =(X max +X min ) / 2;

[0074] Step 4: The weighted probability distribution function pdf obtained in step 3 w(l) performing cumulative calculation to obtain the cumulative distribution function cdf(l), the calculation formula is as follows;

[0075]

[0076] Where, Σpdf w Expressed as the weighted probability distribution function pdf w The cumulative sum of (l); l max Expressed as the weighted probability distribution function pdf w (l) The maximum gray value l in (l);

[0077] The original high-bit deep infrared image I is subjected to weighted gamma correction by the cumulative distribution function cdf(l) to obtain a high-bit deep infrared image T after contrast enhancement, and the calculation formula is as follows:

[0078]

[0079] Where, T n (l) represents the pixel gray value after the weighted gamma correction corresponding to the pixel gray value l;

[0080] Step 5: filtering the original high-bit deep infrared image I to obtain a filtered high-bit deep infrared image I′;

[0081] Step 6: Perform differential calculation on the filtered high-bit deep infrared image I′ to obtain the detail texture information I of the original high-bit deep infrared image I d , the calculation formula is as follows;

[0082] I d (x,y)=I(x,y)-I′(x,y)

[0083] In the formula, x and y represent the position coordinates of the pixel point;

[0084] Step 7: Detail texture information I of the infrared image obtained in step 6 d Perform truncation processing to purify the infrared image detail texture information. At the same time, this step can be omitted in exchange for shorter processing time. The calculation formula is as follows;

[0085]

[0086] In the formula, ε is expressed as a constant, and the size of ε determines the strength of background purification;

[0087] Step 8: The contrast-enhanced high-bit deep infrared image T and the infrared image detail texture information I obtained in step 5 and step 7 are respectively d Perform the weighted probability distribution function pdf obtained in step 3 wThe grayscale mapping of (l) can obtain the displayable infrared image T′ and detail texture information I d ′, the calculation formula is as follows;

[0088]

[0089]

[0090] Where, T max 、T min Respectively expressed as the weighted probability distribution function pdf w (l) Starting from 0 and starting from l max When the traversal starts, the l value is greater than the set parameter δ; I dmax ,I dmin Similarly, parameter δ is an adjustable parameter used to control the brightness of the image after grayscale mapping. This parameter can be an adjustable value t, with δ = t, or an adjustable thousandth of the total number of pixels t‰, with δ = t*W*H‰, for T max and T min The adjustable parameter δ can be taken to different values ​​by adjusting the numerical value t or the thousandth ratio t‰.

[0091] Step 9: The infrared image T′ and the detail texture information I obtained after grayscale mapping in step 8 are d ′ is fused, and the infrared image T′ and the detail texture information I d ′ are all 8-bit images, which can be directly fused and displayed to obtain the enhanced displayable infrared image I e , the fusion method is weighted processing, and the calculation formula is as follows;

[0092] I e (x,y)=λT′(x,y)+γI d ′(x,y)

[0093] Where, parameters λ and γ are adjustable coefficients, λ∈[0,1], γ∈[0,1];

[0094] Example 2

[0095] In this embodiment, the high dynamic infrared digital image detail quality improvement method based on weighted gamma correction proposed in Embodiment 1 is applied to the quality improvement of conventional 8-bit images. The specific operation steps are basically the same as those in Embodiment 1, and the different steps are as follows:

[0096] Step 8, only the infrared image detail texture information I d To do grayscale mapping, the calculation is shown as follows;

[0097]

[0098] In the formula, I dmax ,I dmin Respectively expressed as the weighted probability distribution function pdf w (l) Starting from 0 and starting from l max When the traversal starts, the l value is greater than the set parameter δ; the parameter δ is an adjustable parameter used to control the brightness of the image after grayscale mapping. The parameter can be an adjustable value t, with δ = t, or an adjustable thousandth of the total number of pixels t‰, with δ = t*W*H‰, which is used for I dmax and I dmin The adjustable parameter δ can be taken to different values ​​by adjusting the numerical value t or the thousandth ratio t‰.

Claims

1. A method for improving the detail quality of high dynamic infrared digital images based on weighted gamma correction, characterized in that: The following steps are involved: Step S1, reading an original high-bit depth infrared image I; the bit depth of the original high-bit depth infrared image I is 10 / 12 / 14 / 16 bits, and recording the width W and height H of the original high-bit depth infrared image I; Step S2, performing histogram analysis on the original high-bit deep infrared image I to obtain a probability distribution function pdf(l); Step S3, weighting the probability distribution function pdf(l) to obtain a weighted probability distribution function pdf w (l); Step S4, weighted probability distribution function pdf w (l) performing cumulative calculation to obtain the cumulative distribution function cdf(l); The original high-bit deep infrared image I is weighted gamma corrected using the cumulative distribution function cdf(l) to obtain a high-bit deep infrared image T with enhanced contrast. Step S5, filtering the original high-bit deep infrared image I to obtain a filtered high-bit deep infrared image I′; Step S6, performing differential calculation on the filtered high-bit deep infrared image I′ to obtain the detail texture information I of the original high-bit deep infrared image I d ; Step S8: respectively performing contrast enhancement on the high-bit deep infrared image T and the infrared image detail texture information I d Grayscale mapping is performed to obtain the displayable infrared image T′ and detail texture information I d ′; Step S9, the infrared image T′ and the detail texture information I after grayscale mapping d ′ is fused to obtain the enhanced displayable infrared image I e .

2. The method for improving detail quality of high dynamic infrared digital images according to claim 1, characterized in that: The step S2 further comprises: The probability distribution function pdf(l) is calculated as follows: Where n l Represents the number of pixels with grayscale value l in the original high-bit deep infrared image I.

3. The method for improving detail quality of high dynamic infrared digital images according to claim 1, characterized in that: The step S3 further comprises: Weighted probability distribution function pdf w (l), calculated as follows: In the formula, pdf max 、pdf min They represent the maximum and minimum values ​​of the probability distribution function pdf(l) respectively; α represents the adjustable power coefficient, α>0; β represents the linear coefficient; X M , X G , X max , X min They represent the mean, median, maximum, and minimum grayscale values ​​of the original high-bit deep infrared image I pixel, respectively. The median X G For: X G =(X max +X min ) / 2.

4. The method for improving detail quality of high dynamic infrared digital images according to claim 1, characterized in that: The step S4 further comprises: The cumulative distribution function cdf(l) is calculated as follows: Where, Σpdf w Represents the weighted probability distribution function pdf w The cumulative sum of (l); l max Represents the weighted probability distribution function pdf w The maximum grayscale value l in (l).

5. The method for improving detail quality of high dynamic infrared digital images according to claim 1, characterized in that: The step S4 further comprises: The pixel grayscale value after weighted gamma correction is calculated as follows: Where, T n (l) represents the pixel gray value after weighted gamma correction corresponding to the pixel gray value l.

6. The method for improving detail quality of high dynamic infrared digital images according to claim 1, characterized in that: The step S6 further comprises: Detailed texture information of the original high-bit deep infrared image I d , the calculation formula is as follows; I d (x,y)=I(x,y)-I′(x,y) In the formula, x and y represent the position coordinates of the pixel point.

7. The method for improving detail quality of high dynamic infrared digital images according to claim 1, characterized in that: The step S8 further comprises: Based on the weighted probability distribution function pdf w (l), grayscale mapping is performed, and the calculation formula is as follows: Where, T max , T min Represent the weighted probability distribution function pdf w (l) Starting from 0 and starting from l max When the traversal starts, the l value is greater than the set parameter δ; I dmax ,I dmin Similarly, parameter δ is an adjustable parameter. This parameter is an adjustable value t, δ = t, or it can be an adjustable thousandth of the total number of pixels t‰, δ = t*W*H‰, For T max and T min The adjustable parameter δ can take different values ​​by adjusting the numerical value t or the thousandth ratio t‰.

8. The method for improving detail quality of high dynamic infrared digital images according to claim 1, characterized in that: The step S9 further comprises: The fusion process is weighted processing, and the calculation formula is as follows: I e (x,y)=λT′(x,y)+γI d ′(x,y) Where λ and γ are adjustable coefficients, λ∈[0,1], γ∈[0,1].

9. The method for improving detail quality of high dynamic infrared digital images according to any one of claims 1 to 8, characterized in that: The method further comprises step S7, wherein the infrared image detail texture information I d After truncation, the detailed texture information of the purified infrared image is obtained.

10. The method for improving detail quality of high dynamic infrared digital images according to claim 9, characterized in that: The step S7 further comprises: Truncation processing, the judgment formula is as follows; In the formula, ε is a constant.

Images