Partial Differential Equations-based Image Segmentation Algorithm Research And Application

Image segmentation based on partial differential equations, especially the employment of active contour model, has recently been studied actively and emerged into a key technique in the fields of image processing research.After a brief review of the literature related to image segmentation technique, partial differential equation method in image processing and the active contour model, several active contour algorithms have been discussed, which include: (1) A new geometric active contour algorithm, has been researched. An adaptive construction method of signed distance function has been constructed to reduce the computational payload; (2) An improvement of Chan-Vese algorithm ground on Mumford-Shah model has been performed. By introducing a new energy term which can maintain the level set function as an approximate signed distance function near the zero level set into the model, not only the reinitialization procedure is avoided and much more flexible initialize scheme can be selected, but also the result of segmentation is improved and the curve evolution is speeded up dramatically; (3) A multiphase level set algorithm has been implemented for segmentation of multiple objects; (4) One multiresolution active contour algorithm for video tracking has been implemented; (5) One motion objects detection algorithm has been implemented. The motion boundary is detected via interframe image difference at first, then the contour is extracted using the improved Chan-Vese algorithm, at last a modified region growing scheme has been employed to get more accurate result.Finally, the thesis has pointed out several open problems and further research directions.