Research On Image Background Light Elimination Technology

It is an important to remove effectively the interference of background light in computer vision.The color and texture of the target object will be annihilated under the background light since the signal of the background light is usually much higher than that of the object or its smooth surface can easily bring to specular reflection(highlight),and thus decreases the quality of the image.After eliminating the interference,it can get the more original color and texture information of the image and produce better results for further works.This study focuses on the background light removal,and proposes two background light removing methods based on auxiliary light source imaging and highlight removal.In the process of image acquisition,laser lock-in imaging is an excellent technology for eliminating the effects of the background light by a sine auxiliary modulated laser.However,the technology is expensive and not safe enough,and it requires that the frame rate of the image sensor is high enough to meet the modulation speed.In this study,the technology is extended theoretically to investigate the feasibility of more non-sine auxiliary modulated light and its mechanism.We employed an LED as the auxiliary light source and optimized the control mode of image acquisition to lower the requirement of the frame rate of the image sensor.The study indicates that the signal used to modulate the light source may be any signal whose integral is zero in a period.The experiments based on sine,square,triangle and sawtooth signals show that the four modulations can all achieve good background light elimination performance under different sampling rates on a low frame sensor,which verifies the feasibility of the theory.For a digital image with background highlight,an improved algorithm with joint bilateral filtering is proposed to effectively remove the highlight while preserving more image color and texture information.The algorithm separates firstly the image pixels into two categories,the pixels of only diffuse reflection and those of specular reflection,by an optimized threshold.And then it estimates the maximum diffuse chromaticity of the two categories of pixels respectively.As a result,the procedure improved the estimation precision of the maximum diffuse chromaticity.Finally,let the estimated image with maximum diffuse chromaticity be a joint bilateral filtering guide map to smooth the image with maximum highlight chromaticity,and produce the specular-free image.The experiment results show that,comparing to the existing algorithms,the improved algorithm not only removes the highlights more effectively,but also retains the more color and texture information of the original image.And,the average peak signal-to-noise ratio of the improved algorithm is increased by 3.850 d B while the average structure similarity is increased by 0.083.