The Applications Of Multiscale Geometry Analyse In Edge Detection

The edge is one of the most important features in images, which contains lots of information. The effect of edge detection has a great influence on image segmentation and pattern recognition. The traditional operators detect edge through extracting singular points mostly, these operators search for edge just in some directions limited, which can not use the geometrical property of images.The image processing based on multiscale geometry analyse could make use of the geometrical property of images better. The beamlet and wedgelet transform of multiscale geometry analyse are selected to research the edge detection of images. The major researches of this paper are as follow: the first is to propose one or two new edge detection algorithms based on multiscale geometry analyse; another work is to carry out physical verification on embedded system for new algorithms, and optimize it on embedded system.The main contributions and creative works of the thesis are summarized as follows:(1) A new edge detection algorithm is proposed through improving the algorithm of linear feature extraction based on beamlet transform.(2) Another new edge detection algorithm is proposed based on multiscale wedgelet theory.(3) We build up an embedded system based on DaVinci technology of TI company, using the system to gather, process and display images. The physical verification of new algorithm based on beamlet transform is carryed out on embedded system. The experiment shows that: the new algorithm has good performance on edge extraction, which is also suit for using in engineering.(4) On hardware platform, as the new algorithm has optimized, which have raised the speed of new algorithm greatly.The new algorithms based on beamlet and wedgelet transform both have good performance on edge detection. The edge we gained consist of many beamlets that contains information on the position, the orientation and so on, which will be convenient for the target recognition.