Image coding by adaptive tree-structured segmentation

A new segmentation-based image coding method is proposed. The encoder recursively partitions an image into convex n-gons, 3<or=n<or=8, until the pixels satisfy a uniformity criterion. The partition is adaptive to image semantics by aligning these n-gons with image edges. The compressed image is sent as a labeled pointerless binary tree, and decoding is simply polygon filling. The adaptability of the partition is achieved by an optimized 2-dimensional piecewise constant approximation of the image. This seemingly expensive optimization process is made computationally feasible by a novel preprocessing technique. High compression ratios are obtained through a good compromise between the segmentation validity and segmentation complexity.<<ETX>>