Wavelet based multiscale edge preserving segmentation algorithm for object recognition and object tracking

Although edge detection is basic task in computer vision, its results are very important for further procedures. In order to make task of finding edges of objects more efficient, a novel approach is proposed. It uses wavelet transform of the image to detect edges and in the same time also to segment edge points. Partial segmentation of the image is performed only around edge points, which in return enables us to preserve found edges, group edges of the same object into clusters and defines shapes of objects, where edges are missing. Because we do not perform segmentation of the whole image, but use only information from segmentation of point around edges, the procedure of finding object shapes is faster and in turn the whole step of object recognition, defined by the shape, is faster.