System Of Contour And Sketch Extraction From Video Sequence Frame Based On Manual Template

Contour extraction is an important pre-processing of image processing, computer vision, pattern recognition, and it has wide applications in visual matching, motion tracking, 3D reconstruction, image recognition, etc. Sketch extraction which is developed on the basis of the contour extraction technology has become an important method of video and film post-production effects, and it is one of the most popular areas in computer vision.In this thesis, we implement an interactive sketch and contour extraction framework of integrating boundary detection, shape matching, thin-plate spline and snake model. Given an input video, this system can extract object's contour or sketch interactively according to the model.Our work is mainly about:(1) Based on analyses of present research on edge detection, we explain the primal sketch algorithm, and discuss the algorithmic process through some sequence frame and image experiments.(2) Considering the objects between adjacent frames have small deformation, together with the advantage of Shape Tree Matching algorithm, we design a matching algorithm between templates and sketch. Some experiments are performed and the results are analyzed.(3) We study the thin-plate spline and design a template transform algorithm based on TPS, discuss the problems in application through experiments.(4) Since snake model is sensitive to its initialize position, we design the contour smoothing algorithm based TPS with active contour model.