An adaptive fusion method for images with large linear dynamic range

Abstract

The present invention provides a method for adaptive fusion of images with a large linear dynamic range to address the current problem of fusing two images with different gains to obtain a high-quality image with a large linear dynamic range. This method requires extensive pre-calibration of the imaging system's high- and low-gain ratios and background values before fusion. Furthermore, changes in the external environment can cause these ratios and background values to change, leading to errors or faults in the fused result after calibration. The present invention classifies the pixels in the two acquired high- and low-gain images, calculates the imaging system's true high- and low-gain ratios, and calculates the compensated background during fusion, thereby achieving adaptive fusion of images with a large linear dynamic range. No pre-calibration is required for fusion of images with a large linear dynamic range, effectively preventing the influence of the external environment on the gain ratio and background.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a method for adaptive fusion of images with a large linear dynamic range. Background Art

[0002] In research fields such as biological imaging, biomedicine, and high-energy physics, the target images to be imaged often have a large dynamic range, which requires the imaging system to also have a large linear dynamic range. Currently, dual-gain image sensors or multiple exposure techniques are used to fuse high- and low-gain images to obtain images with a large linear dynamic range. The perfect fusion of high- and low-gain images is crucial for obtaining high-quality images with a large linear dynamic range.

[0003] Currently, there are many methods for fusing high- and low-gain images, but all require pre-calibration of the imaging system's high- and low-gain ratio and background value, which requires a lot of preliminary work. If the external environment changes during the fusion process, the high- and low-gain ratio and background value will change, resulting in a poor fusion effect or even erroneous fusion results. Summary of the Invention

[0004] The purpose of the present invention is to solve the technical problem that the current method of obtaining a high-quality image with a large linear dynamic range by fusing two images with different gains requires a lot of preliminary calibration work on the high and low gain ratios and background values of the imaging system in the early stage of fusion, and when the external environment changes, the high and low gain ratios and background values will change, resulting in errors or faults in the fused result after calibration. The present invention provides an adaptive fusion method for images with a large linear dynamic range.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is:

[0006] A large linear dynamic range image adaptive fusion method, which is special in that it includes the following steps:

[0007] Step 1: Acquire two gain images by multiple exposures or simultaneous exposures, where the high-gain image is H and the low-gain image is L, and record the pixel grayscale values of all pixels in the two gain images respectively;

[0008] Step 2: Create pixel sets SH and SL for the high-gain image H and the low-gain image L, respectively, set the linear range of pixel grayscale values from a to b, and assign pixels with grayscale values at the same position in the high-gain image H and the low-gain image L within the linear range of pixel grayscale values from a to b to the pixel sets SH and SL respectively;

[0009] Step 3: Based on the grayscale values of the two gain images, set the classification rules and establish subsets SH1, SH2 and SL1, SL2 of the pixel sets SH and SL respectively, and classify the pixels in the pixel sets SH and SL again;

[0010] Step 4: Calculate the average grayscale value of the pixels in the pixel sets SH and SL, and the subsets SH1, SH2 and SL1, SL2 respectively;

[0011] Step 5: Calculate the gain ratio G and fusion compensation background value V of the two images b ;

[0012] G=(E SH1 -E SH2 ) / (E SL1 -E SL2 )

[0013] Among them, E SH1 、E SH2 、E SL1 and E SL2 are the average grayscale values of pixels in subsets SH1, SH2, SL1, and SL2 respectively;

[0014] V b =E SH -G×E SL

[0015] Among them, E SH and E SL are the average grayscale values of the pixels in the pixel sets SH and SL respectively;

[0016] Step 6: Traverse all pixels in the high-gain image H and the low-gain image L and fuse them to obtain a fused HDR image with a large linear dynamic range.

[0017] Furthermore, in step 3, the classification rules are set to classify the pixels in the pixel sets SH and SL again, specifically:

[0018] The pixel points with grayscale values greater than the median in the pixel set SH are classified into the subset SH1, and the rest are classified into the subset SH2; the pixel points with grayscale values greater than the median in the pixel set SL are classified into the subset SL1, and the rest are classified into the subset SL2.

[0019] Furthermore, step 6 is specifically as follows:

[0020] Traverse all the pixels in the high-gain image H and the low-gain image L for fusion. When the gray value H of the pixel (i, j) in the high-gain image H is i,jWhen it is greater than b, the pixel grayscale value HDR of the pixel in the fused large linear dynamic range HDR image is assigned i,j =G×L i,j +V b Otherwise, the pixel gray value HDR of the pixel in the fused large linear dynamic range HDR image is assigned i,j =H i,j , get the fused large linear dynamic range HDR image;

[0021] Wherein, 0≤i≤m-1, 0≤j≤n-1; m is the number of pixel rows of the high-gain image H and the low-gain image L, and n is the number of pixel columns of the high-gain image H and the low-gain image L.

[0022] Furthermore, in step 1, the obtaining of two gain images by multiple exposures or simultaneous exposures is specifically to adopt an imaging system with high and low dual gain channel outputs to obtain two gain images by multiple exposures or simultaneous exposures.

[0023] Furthermore, in step 2, a is the minimum value of the high-gain output linear range of the imaging system described in step 1, and b is the maximum value of the high-gain output linear range of the imaging system.

[0024] Compared with the prior art, the present invention has the following beneficial technical effects:

[0025] The adaptive fusion method for large linear dynamic range images, provided by this invention, classifies pixels in two acquired high- and low-gain images to calculate the imaging system's true high- and low-gain ratio, as well as the compensation background during fusion. This method achieves adaptive fusion of large linear dynamic range images. Pre-calibration is not required for large linear dynamic range image fusion, effectively minimizing the effects of the external environment on the gain ratio and background. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is a flow chart of the large linear dynamic range image adaptive fusion method of the present invention;

[0027] Figure 2 Schematic diagram of the imaging system structure with high and low dual gain channel outputs used to obtain gain images in this embodiment. DETAILED DESCRIPTION

[0028] To further clarify the objectives, advantages, and features of the present invention, the following describes in further detail the method for adaptive fusion of images with a large linear dynamic range proposed by the present invention, with reference to the accompanying drawings and specific examples. Those skilled in the art should understand that these embodiments are intended only to illustrate the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0029] like Figure 1 As shown, the large linear dynamic range image adaptive fusion method provided by this embodiment specifically includes the following steps:

[0030] Step 1: Acquire two gain images by multiple exposures or simultaneous exposures, where the high-gain image is H and the low-gain image is L, and record the pixel grayscale values of all pixels in the two gain images respectively;

[0031] Let the number of pixel rows of the two gain images be m, the number of pixel columns be n, and the pixel grayscale values of the pixel point (i, j) in the gain image be expressed as H i,j 、L i,j , where 0≤i≤m-1, 0≤j≤n-1;

[0032] When acquiring gain images, an imaging system with high and low dual gain channel outputs can be selected (such as Figure 2 As shown), a high-gain image is obtained as H and a low-gain image is obtained as L;

[0033] Step 2: Create pixel sets SH and SL for the high-gain image H and the low-gain image L, respectively, set the linear range of pixel grayscale values from a to b, and assign pixels with grayscale values at the same position in the high-gain image H and the low-gain image L within the linear range of pixel grayscale values from a to b to the pixel sets SH and SL respectively;

[0034] a is the minimum value of the high-gain output linear range of the imaging system used in step 1, and b is the maximum value of the high-gain output linear range of the imaging system.

[0035] Step 3: Based on the grayscale values of the two gain images, set the classification rules and establish subsets SH1, SH2 and SL1, SL2 of the pixel sets SH and SL respectively, and classify the pixels in the pixel sets SH and SL again;

[0036] In this embodiment, the classification rule is set based on the median of the grayscale values of the two gain images, that is, the pixels in the pixel set SH whose grayscale values are greater than the median are classified into the subset SH1, and the rest are classified into the subset SH2; the pixels in the pixel set SL whose grayscale values are greater than the median are classified into the subset SL1, and the rest are classified into the subset SL2;

[0037] Alternatively, the pixels in the pixel sets SH and SL may be classified again based on the mean values of the grayscale values of the pixel sets SH and SL.

[0038] Step 4: Calculate the average grayscale value of the pixels in the pixel sets SH and SL, and the subsets SH1, SH2 and SL1, SL2, respectively, and record them as E SH 、E SL 、E SH1 、ESH2 、E SL1 、E SL2 ;

[0039] Step 5: Calculate the gain ratio G and fusion compensation background value V of the two images b ;

[0040] Gain ratio G: G=(E SH1 -E SH2 ) / (E SL1 -E SL2 ),

[0041] Fusion compensation background value V b :V b =E SH -G×E SL ;

[0042] Step 6: Traverse all pixels in the high-gain image H and the low-gain image L to fuse them to obtain a fused HDR image with a large linear dynamic range;

[0043] The fusion process is as follows: when the gray value H of the pixel (i, j) in the high gain image H i,j When it is greater than b, the pixel grayscale value HDR of the pixel in the fused large linear dynamic range HDR image is assigned i,j =G×L i,j +V b Otherwise, the pixel gray value HDR of the pixel in the fused large linear dynamic range HDR image is assigned i,j =H i,j , that is: H i,j >b, HDR i,j =G×L i,j +V b ;H i,j When ≤b, HDR i,j =H i,j .

[0044] The present invention utilizes an imaging system with dual high- and low-gain channel outputs to capture two high- and low-gain images. By classifying the pixels in the two gain images, the imaging system's true high- and low-gain ratio, as well as the compensation background during fusion, is calculated, enabling adaptive fusion of images with a large linear dynamic range. This fusion of large linear dynamic range images does not require pre-calibration, effectively preventing the effects of the external environment on the gain ratio and background.

[0045] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the present invention.

Claims

1. A large linear dynamic range image adaptive fusion method, characterized in that: The following steps are involved: Step 1: Acquire two gain images by multiple exposures or simultaneous exposures, where the high-gain image is H and the low-gain image is L, and record the pixel grayscale values of all pixels in the two gain images respectively; Step 2: Create pixel sets SH and SL for the high-gain image H and the low-gain image L, respectively, set the linear range of pixel grayscale values from a to b, and assign pixels with grayscale values at the same position in the high-gain image H and the low-gain image L within the linear range of pixel grayscale values from a to b to the pixel sets SH and SL respectively; Step 3: Based on the grayscale values of the two gain images, set the classification rules and establish subsets SH1, SH2 and SL1, SL2 of the pixel sets SH and SL respectively, and classify the pixels in the pixel sets SH and SL again, specifically: The pixels in the pixel set SH whose grayscale values are greater than the median are grouped into the subset SH1, and the rest are grouped into the subset SH2; the pixels in the pixel set SL whose grayscale values are greater than the median are grouped into the subset SL1, and the rest are grouped into the subset SL2; Step 4: Calculate the average grayscale value of the pixels in the pixel sets SH and SL, and the subsets SH1, SH2 and SL1, SL2 respectively; Step 5: Calculate the gain ratio G and fusion compensation background value V of the two images b ; G=(E SH1 -E SH2 ) / (E SL1 -E SL2 ) Among them, E SH1 、E SH2 、E SL1 and E SL2 are the average grayscale values of pixels in subsets SH1, SH2, SL1, and SL2 respectively; V b =E SH -G×E SL Among them, E SH and E SL are the average grayscale values of the pixels in the pixel sets SH and SL respectively; Step 6: Traverse all pixels in the high-gain image H and the low-gain image L and fuse them to obtain a fused HDR image with a large linear dynamic range. Specifically: Traverse all the pixels in the high-gain image H and the low-gain image L for fusion. When the gray value H of the pixel (i, j) in the high-gain image H is i,j When it is greater than b, the pixel grayscale value HDR of the pixel in the fused large linear dynamic range HDR image is assigned i,j =G×L i,j +V b Otherwise, the pixel gray value HDR of the pixel in the fused large linear dynamic range HDR image is assigned i,j =H i,j , get the fused large linear dynamic range HDR image; Where, 0≤i≤m-1, 0≤j≤n-1; m is the number of pixel rows of the high-gain image H and the low-gain image L, n is the number of pixel columns of the high-gain image H and the low-gain image L; L i,j is the gray value of pixel (i, j) in the low-gain image L; V b Compensate for background values for fusion.

2. The large linear dynamic range image adaptive fusion method according to claim 1, characterized in that: In step 1, the obtaining of two gain images by multiple exposures or simultaneous exposures is specifically to use an imaging system with high and low dual gain channel outputs to obtain two gain images by multiple exposures or simultaneous exposures.

3. The large linear dynamic range image adaptive fusion method according to claim 2, characterized in that: In step 2, a is the minimum value of the high-gain output linear range of the imaging system described in step 1, and b is the maximum value of the high-gain output linear range of the imaging system.

Images