Research Of Multiresolution-Based Watersheds For Image Segmentation

Image segmentation is a key technology of image processing, and its purpose is separating the target from the background. It makes further image analysis and comprehension possible. There are a lot of classical image segmentation algorithms already, but no one has good result for all kinds of image. To improve the quality of image segmentation, we should carry on the research based on characteristic of a given image, using the priori knowledge, and focused on specified segmentation modal.The watershed is a kind of mathematical morphologic image segmentation. It get the precise edge which is continues, closing and single-pixel. The main limitation of watershed transform is the over-segmentation due to its sensitive to noise; even the very thin noise will lead to a lot of scattered and meaningless regions. To overcome over-segmentation, in this paper, we adopt the multiresolution-based watersheds, and get the following results:1) We reduce the noise by using the wavelet transform and do image segmentation on low-solution image which has much smaller size than the original and computational complexity is strongly reduced.2) Choosing the merging discipline for different kind of image, so it leads to the homogeneous regions in the segmentation result.3) We modify the segmentation result of high-solution image by that of lower-solution image, and this avoids the problem of sawtooth edge.4) Compared with the method of first watershed segmentation and then merging regions, the method of watershed multiresolution-based shows better resultand the speed is faster.