US 11,989,850 B2
Method for calculating image highlight reflection separation based on polarization guidance
Feng Lu, Beijing (CN); and Sijia Wen, Beijing (CN)
Assigned to Beihang University, Beijing (CN)
Filed by Beihang University, Beijing (CN)
Filed on Dec. 31, 2021, as Appl. No. 17/566,858.
Claims priority of application No. 202110076538.0 (CN), filed on Jan. 20, 2021.
Prior Publication US 2022/0230273 A1, Jul. 21, 2022
Int. Cl. G06T 3/04 (2024.01); G06T 7/557 (2017.01); G06T 7/593 (2017.01)
CPC G06T 3/04 (2024.01) [G06T 7/557 (2017.01); G06T 7/593 (2017.01)] 5 Claims
OG exemplary drawing
 
1. A method for calculating image highlight reflection separation, comprising:
controlling a polarized color sensor to shoot a polarized color image, and calculating a polarized chromaticity image according to polarization information and color information of the polarized color image;
based on characteristics of the polarization information, using the polarized chromaticity image as a guide, clustering pixels of diffuse reflection components whose chromaticity differences are within a preset range to obtain a pixel sequence set, wherein the characteristics of the polarization information are the characteristics that, in case the polarized chromaticity image has highlight information, color information of the diffuse reflection components of the polarized color image is still obtained, and wherein each pixel in every pixel sequence in the pixel sequence set has a diffuse reflection component whose chromaticity meets a preset condition, and due to difference in surface material and lighting direction, each pixel in every pixel sequence in the pixel sequence set has a different degree of highlight reflection component;
using a robust principal component analysis algorithm for each pixel in every pixel sequence in the pixel sequence set, to retain the diffuse reflection component of each pixel in the pixel sequence, and at the same time, remove the highlight reflection component of each pixel in the pixel sequence, and in the end, preliminarily separate the polarized chromaticity image into the diffuse reflection component and the highlight reflection component, to obtain an initial separation result, wherein the initial separation result includes the diffuse reflection component and the highlight reflection component, and there is an invariant highlight reflection component in the diffuse reflection component;
using the initial separation result as a constraint and based on an iterative optimization algorithm of an alternating direction method of multipliers, further processing the polarized color image for highlight reflection separation to obtain a highlight reflection separation result, and generating a diffuse reflection image based on the highlight reflection separation result.