Research On Specular Reflection Elimination Based On Image Rendering

The intensity of pixels,in the optical digital image,is related to the object reflection.The characteristics of the object are reflected by the diffuse reflection components,and the characteristics of the light source are reflected by the specular reflection components.When the intensity of the specular reflection components are greater than the upper limit of the acquisition device,the intensity of pixels is saturated,and the object information contained in the pixel is lost.If the image contained saturated pixels is directly matched,split,etc.,the processing result will be inaccurate.Therefore,it is necessary to eliminate the specular reflection components in the image,restore the lost information,and generate a high quality image for subsequent processing.Image rendering technology is often used to render realistic images.The images under arbitrary illumination conditions are generated through a set of image sequences with different illumination conditions by image-based rendering technology.Therefore,the specular reflection components are eliminated by generating the diffuse reflection components of the image.In this dissertation,the technology of the specular reflection elimination based on image rendering is mainly studied,under varied illumination,fixed camera and scene.In this dissertation,the commonly used specular reflection elimination algorithms are introduced,the characteristics and applicable conditions of these algorithms are analyzed and summarized,and then the methods of photometric correction and photometric linearization are introduced;the linear relationship of diffuse reflectance components based on geometrical optics is deduced,and the conclusion that the linear combination coefficients of diffuse reflectance components in the neighborhood are approximate under the condition of near light source is obtained;the maximum chromaticity constancy of homogeneous pixels is verified,which provides a basis for excluding the pixels containing specular reflectance components by detecting abnormal values.inspired by the existing methods and deduction conclusions,this paper proposes two methods to generate linearized images based on random sampling and neighborhood.The way for detecting outliers of both methods is Grubbs detection.According to the synthetic image results,the neighborhood-based method is selected to generate the linearized image in the proposed specular reflection elimination algorithm.The experimental results show that the proposed method is superior to the existing photometric rendering technology with 9 images,and the specular reflection components are eliminated effectively,when applied to objects with complex geometric properties.