CN-121981928-A - Dynamic range compressed image enhancement algorithm for ultrahigh resolution high dynamic image and hardware thereof

CN121981928ACN 121981928 ACN121981928 ACN 121981928ACN-121981928-A

Abstract

The invention discloses a dynamic range compressed image enhancement algorithm for an ultrahigh resolution high dynamic image and hardware thereof, wherein the algorithm comprises the steps of RGB-XYZ color space conversion, n-frame average maximum brightness statistics, self-adaptive brightness scaling, dynamic mapping factor calculation, segmentation low bit mapping and the like, realizes high dynamic range accurate compression, and the hardware adopts a modularized design to support ASIC/FPGA deployment. The method and the device can efficiently complete the dynamic range compression of the high-bit high-dynamic image to 8-bit display, reserve the details of dark and medium-brightness areas to the maximum extent, restore the visual effect of the original scene, abandon the traditional high-complexity algorithm, reduce the operand and the on-chip storage cost by simplifying the design, realize low-delay low-power deployment, avoid the overexposure and underexposure caused by fixed mapping curves by adopting self-adaptive parameters and dynamic mapping, remarkably improve the quality of the reconstructed image, and perfectly adapt to the high-quality display requirement under the low-delay low-power scene.

Inventors

WANG SHIHAO
ZHANG BIN
HUANG JINYUAN
JI WEN
YAO JUN

Assignees

中科工业人工智能研究院

Dates

Publication Date: 20260505
Application Date: 20251202

Claims (8)

1. A dynamic range compressed image enhancement algorithm for ultrahigh resolution high dynamic images comprises the following steps: S1, converting an input high-dynamic RGB image pixel into an XYZ color space, acquiring X, Y, Z components, taking a Y component as the world brightness of a current scene, counting the maximum world brightness Ymax of the current frame, repeating the counting process for n frames, calculating the average maximum world brightness Ymax_avg of the current scene, and setting the maximum brightness of a screen to be Ydmax; s2, carrying out normalization processing on the obtained X, Y components to obtain normalized X components and normalized Y components; S3, setting a brightness scaling parameter b according to the brightness feeling of the current scene, and scaling all Y components of the current frame to obtain a brightness scaling intermediate value Yscale; s4, mapping the obtained Yscale to a logarithmic domain to obtain a logarithmic domain brightness value Ylog; s5, calculating a mapping factor D of the pixel at the current position by utilizing Yscale, the average maximum world brightness Ymax_avg and the brightness scaling parameter b; S6, calculating brightness Ynew of the pixel at the current position after the dynamic range is compressed according to the parameters obtained in the steps S1 to S5; s7, according to Ynew, xnew and Znew are calculated, and Xnew, ynew, znew is remapped to an RGB domain; S8, according to the brightness characteristics of the RGB pixels after dynamic range compression, mapping the RGB pixels to the corresponding low-bit pixels in a segmented mode.
2. The method for dynamic range compression image enhancement for ultrahigh resolution high dynamic image according to claim 1, wherein in step S1, the formula for converting RGB image pixels into XYZ color space is: ; ; ; Wherein x0, x1, x2, y0, y1, y2, z0, z1, z2 are scaled conversion factors, bitdepth are bit widths of the input high-dynamic image; The calculation formula of ymax_avg is: ; Where p is the current frame number.
3. The dynamic range compressed image enhancement algorithm for ultrahigh resolution high dynamic images according to claim 2, wherein in step S2, the normalization processing is performed in the following specific manner: ; If sum_xyz >0, xx=x/sum_xyz, yy=y/sum_xyz, and if sum_xyz=0, xx=xx_normal, yy=yy_normal, where sum_xyz is X, Y, Z components Sum, xx and yy are normalized X, Y components, respectively, and xx_normal and yy_normal are default values to avoid overflow.
4. The method for enhancing a dynamic range compressed image for ultrahigh resolution high dynamic image according to claim 3, wherein in step S3, the calculation formula of luminance scaling is: ; and the brightness scaling parameter b is positively correlated with the subjective brightness of the current scene, and eps in the formula is a self-defined brightness fuzzy factor.
5. The method for dynamic range compression image enhancement for ultrahigh resolution high dynamic image according to claim 4, wherein in step S4, the formula mapped to the logarithmic domain is: ; wherein k is a logarithmic domain mapping parameter; in step S5, the calculation formula of the dynamic mapping factor D is: ; wherein D0 and D1 are scaling parameters of the dynamic mapping factor D; in step S6, the calculation formula of Ynew is: ; ; ; Where ymax_avg_factor is the average maximum luminance ratio, ymax_log10 is the log domain map approximation, and α is the average maximum luminance scaling parameter.
6. The dynamic range compressed image enhancement algorithm for ultrahigh resolution high dynamic images according to claim 5, wherein in step S7, xnew and Znew are calculated by: Xnew= Ynew/yyxxx, znew = Ynew/(1-xx-yy) if yy >0, xnew=0, znew=0 if yy is less than or equal to 0; xnew, ynew, znew the formula mapped to the RGB domain is: ; ; ; wherein Ynew is a target luminance value after compression, xnew and Znew are chromaticity components after compression X, Z, respectively, and r0, r1, r2, b0, b1, b2, g0, g1 and g2 are conversion parameters after scaling.
7. The dynamic range compressed image enhancement algorithm for ultrahigh resolution high dynamic images according to claim 6, wherein in step S8, the specific way of segment mapping is: G% of peak pixel value is taken as a boundary line; If pixel < g%, then pixel_out=pixel×β, if pixel is not less than g%, then pixel_out=γ× (pixel δ) - λ; wherein g is a customizable mapping demarcation value, beta is a linear mapping parameter, and gamma, delta and lambda are customizable nonlinear mapping parameters; The mapping formula of the low bit pixel is: ; wherein pixel is a compressed RGB pixel value, pixel_out is a segment map normalized value, bitdepth is 8 bits, and pixel_8bit is a low bit output pixel value.
8. Hardware for realizing the dynamic range compressed image enhancement algorithm according to any one of claims 1 to 7, which is characterized by comprising an RGB-XYZ conversion module, an XY normalization module, a brightness scaling module, a mapping factor searching and calculating module, an XZ component operation module, an XYZ-RGB inverse conversion module and an RGB segmentation mapping module; The RGB-XYZ conversion module is used for receiving 20bit/16bit/12bit/10bit high-dynamic RGB pixels in ISPPIPELINE streams; the XY normalization module is used for normalizing and outputting a X, Y component after normalization; the brightness scaling module is used for inputting a brightness scaling parameter b, executing brightness scaling calculation and outputting Yscale; the mapping factor searching and calculating module pre-stores logarithmic operation data, and is used for obtaining a dynamic mapping factor D through Yscale table searching and calculating Ynew; The XZ component operation module is used for receiving the X, Y components and Ynew after normalization, executing Xnew and Znew calculation and outputting Xnew and Znew; the XYZ-RGB reverse conversion module is used for executing XYZ-RGB mapping and outputting RGB normalized data after dynamic range compression; the RGB segmentation mapping module is used for executing segmentation mapping and outputting low bit RGB pixels.

Description

Dynamic range compressed image enhancement algorithm for ultrahigh resolution high dynamic image and hardware thereof Technical Field The invention relates to the technical field of image processing, in particular to a dynamic range compressed image enhancement algorithm for an ultrahigh resolution high dynamic image and hardware thereof. Background The image which can be accepted by human eyes has a very large dynamic range, and the image acquisition processing equipment utilizes a high dynamic imaging technology to ensure the integrity of image acquisition information to the maximum extent. However, the development speed of the existing screen display technology cannot be synchronous with the development speed, and most display devices only display 8bit images at present, so that a special algorithm is required to process high bit image data, and the visual effect of an original scene in human eyes can be maintained on an 8bit display. Most of the existing dynamic range compression algorithm technologies use histogram equalization, laplace filtering or multi-scale decomposition and other algorithms to process high-dynamic images, and although the image processing effect is good, the traditional algorithms have high operation complexity and on-chip storage cost, the algorithm time delay is too high, and the implementation is difficult under the scene of low time delay and low power consumption. Most of the traditional global tone mapping algorithms adopt fixed mapping curves, the realization is simpler, but the image characteristics of different application scenes and the situation that the high-dynamic image simultaneously has overexposed and underexposed areas on low-bit display equipment are not considered, so that the quality of the reconstructed image is poor. Disclosure of Invention The present invention aims to overcome or at least partially solve the above problems, and proposes a dynamic range compressed image enhancement algorithm for ultrahigh resolution high dynamic images and hardware thereof. In order to achieve the purpose, the invention adopts the following technical scheme that the dynamic range compressed image enhancement algorithm for the ultrahigh resolution high dynamic image comprises the following steps: S1, converting an input high-dynamic RGB image pixel into an XYZ color space, acquiring X, Y, Z components, taking a Y component as the world brightness of a current scene, counting the maximum world brightness Ymax of the current frame, repeating the counting process for n frames, calculating the average maximum world brightness Ymax_avg of the current scene, and setting the maximum brightness of a screen to be Ydmax; s2, carrying out normalization processing on the obtained X, Y components to obtain normalized X components and normalized Y components; S3, setting a brightness scaling parameter b according to the brightness feeling of the current scene, and scaling all Y components of the current frame to obtain a brightness scaling intermediate value Yscale; s4, mapping the obtained Yscale to a logarithmic domain to obtain a logarithmic domain brightness value Ylog; s5, calculating a mapping factor D of the pixel at the current position by utilizing Yscale, the average maximum world brightness Ymax_avg and the brightness scaling parameter b; S6, calculating brightness Ynew of the pixel at the current position after the dynamic range is compressed according to the parameters obtained in the steps S1 to S5; s7, according to Ynew, xnew and Znew are calculated, and Xnew, ynew, znew is remapped to an RGB domain; S8, according to the brightness characteristics of the RGB pixels after dynamic range compression, mapping the RGB pixels to the corresponding low-bit pixels in a segmented mode. In a preferred embodiment, in step S1, the formula for converting RGB image pixels into XYZ color space is: Wherein x0, x1, x2, y0, y1, y2, z0, z1, z2 are scaled conversion factors, bitdepth are bit widths of the input high-dynamic image; The calculation formula of ymax_avg is: Where p is the current frame number. In a preferred embodiment, in step S2, the normalization process is performed in the following manner: If sum_xyz >0, xx=x/sum_xyz, yy=y/sum_xyz, and if sum_xyz=0, xx=xx_normal, yy=yy_normal, where sum_xyz is X, Y, Z components Sum, xx and yy are normalized X, Y components, respectively, and xx_normal and yy_normal are default values to avoid overflow. In a preferred embodiment, in step S3, the calculation formula of the luminance scaling is: and the brightness scaling parameter b is positively correlated with the subjective brightness of the current scene, and eps in the formula is a self-defined brightness fuzzy factor. In a preferred embodiment, in step S4, the formula mapped to the log domain is: wherein k is a logarithmic domain mapping parameter; in step S5, the calculation formula of the dynamic mapping factor D is: wherein D0 and D1 are scaling parameters of the dynamic mapping factor