GPU Based Contour Interpolation in 3D Image Rendering
Zhangwen Wu · Zhongguo yixue wulixue zazhi · 2012
Objective:As traditional contour interpolation algorithms based on distance map are of low efficiency on CPU,this study brings up an accelerated method based on distance map with best advantage of parallel computing GPU.Methods:shape based interpolation follows these steps:(1) translate to align the two original contours;(2) build a markmap for each original contour;(3) execute distance tranformation and draw two distancemaps;(4) peform linear interpolation between two distancemaps;(5) extract contours from interpolated distancemaps;(6) reverse the contours backto supposed positions.Transfer the distance transforming on to GPU,each GPU thread computes one pixel of distance map seperately:read a mark from markmap as well as a coodinate from contour map,calculate the minimal Euclidean Distance,and save the distance by the mark as a pixel of the distancemap.Results:A best acceleration over 600 times is aquaired in experiments.Conclusions:This method fulfills an obvious acceleration and will accomplish real-time interaction in practical applications.