Font Size: a A A

Research On Video Image Stitching Technique In Real-time System

Posted on:2012-04-30Degree:MasterType:Thesis
Country:ChinaCandidate:Q RuanFull Text:PDF
GTID:2218330362456323Subject:Control theory and control engineering
Abstract/Summary:PDF Full Text Request
Visual information is the straightest way for people to get information from the objective world.As a result of it,some relevant techniques about image and video processing become more and more important in many domains such as virtual reality,three-dimensinal reconstruction,video surveillance,video retrieval, medical image analysis.Because mosaic techniques of video and image can not only broaden human horizon but also eliminate redundancy without resolution reduction,so that they are widely used in these domains. This paper just uses two common usb cameras to simulate vehicle video system ,finally making videos are seamless spliced and real-time displayed.To realize these functions should encompass some steps,which are camera demarcation, distortion correction,feature point extraction and registration,image fusion and video displaying.This paper studys and analyzes the principal techniques which are involved in these steps and finally achieve the expected result.By analyzing several means of calibration,this paper chooses a series of functions provided by OpenCV library to calibrate cameras;Once internal,external and distortion parameters are calculated,images which are about to be spliced could be corrected;Then techniques of image registration are emphatically studied and this paper select methods based on registration of feature point at the end;How to structure Sift and Surf descriptor are described in detail;We would fully prove that Surf is much better than Sift at algorithmic complexity,running speed,stability under several transformations.Then bbf algorithm is used to registrate feature points,and ransac algorithm is used to eliminate the wrong pairs.On the other hand,accurate feature points could be caculated to get transformational matrix between images.In order to make video pictures fused seamlessly,mean value method is compared with weighting method through experiment and this paper presents an improved way by center weighting both at rows and cols to realize seamless mosaic.In the end, DirectShow techniques are utilized to make sure that video pictures could be displayed in time.
Keywords/Search Tags:video and image mosaic, feature point extraction, image registration, seamless mosaic
PDF Full Text Request
Related items