视频运动目标检测及其在视频编码中的应用-信息与通信系统专业论文.docxVIP

浙江大学硕士学位论文Abstract 浙江大学硕士学位论文 Abstract Video motion detection and video compression are the most two important problems in smart surveillance system．Video motion detection Can give alert on unexpected events being detected．Applying motion detection to video coding will emphasize the importance of motion areas and greatly improve the coding efficiency of background areas． This thesis focuses on the research of motion detection and its application in video compression．In motion detection,we firSt introduce some representative approaches，and analyze the corresponding advantages and shortcomings．Under the existed framework，herein we introduce a new approach．According to the spatial siIllilarity in one image，we treat pixels as a whole motion object埘t11 homogeneous intensity．Herein we segment images、析th rainfalling watershed transforming．SO as to improve the operated object from pixel level to region level．Frame difference change detection based on statistical model iS used to suppress the noise and get out of the dilemma of threshold decision．Based on the combination of the spatial and temporal similarity,we construct spatial，temporal，moving-or-not energy functions，apply Markov Random Field algorithm to label the last background and foreground classification． The amount of surveillance video data iS terribly numerous．But meaningful data among it iS much less．Motion areas are always the focus of attention in many situations，and the bitrate of background areas especially when they are noised cannot provide much useful information．In this case．1ess resource would be allocated to transmit and store the background areas in video frames．Based on above analysis．we apply the motion detection in video compression to reduce the bitrate，and alSO Can generate all index to search the events which have happened in surveillance video．It greatly improves the humanity—operations when browsing the video data．In this thesis， motion detection is used to guide the mode decision for every mac