Research And Application Of Extracting Surface Contours Based On Improved Photometric Stereo

With the fast development of digitization and intelligence in various field,more and more scenes need to extract the geometric contour of the surface.Photometric stereo is favored for its simplicity,efficiency and accuracy.However,specular or highlights can cause deviations in the surface normal calculation,which in turn seriously affects the surface contour.In addition,as noise on different albedo regions has different influence,the boundary effect appears in the contour image,which affects the contour extraction.To this end,the paper proposes an improved photometric stereo,which can effectively remove specular reflection,eliminate the boundary effect caused by noise on small albedo regions,and accurately extract the surface contour of the object,which has important practical application value.The algorithm is divided into three major steps:Firstly,the pixel values of the input image are corrected by camera response calibration and light field calibration,so that the image can represent the true irradiance under the ideal light source.Then,the specular reflection component is removed using a highlight removal algorithm,and the surface contours is extracted by photometric stereo.Finally,the contour image is filtered by an improved bilateral filtering algorithm to eliminate the boundary effect and further improve the accuracy of contours extraction.The main contribution of the paper is:(1)A highlight detection algorithm based on pixel chrominance ratio is proposed,which can effectively distinguish the highlight pixels of input image in PS;(2)A highlight removal algorithm based on pseudo-pixel brightness ratio is proposed,which can effectively remove specular reflection by using brightness constraints between images and neighborhood space constraints.As a result,the impact of highlights on contour extraction is greatly reduced.(3)An adaptive Gaussian kernel bilateral filtering algorithm based on albedo is proposed to eliminate the boundary effect on contours image caused by noise in the low albedo regions.The algorithm proposed in this paper has achieved good results on both synthetic images and physical image datasets with high light,which can be effectively applied to code character extraction and surface defect detection.Therefore,the algorithm is a universal method of extracting surface contours.