中文摘要：

提出了一种新的基于分形编码的图像分割算法,其工作原理和实现方法与现有的基于分形编码的图像分割方法完全不同。新算法利用不同尺度间图像块的自相似性,根据编码的结果为不同像素之间定义生成关系,通过选择合适的误差度量使满足生成关系的各像素灰度值足够接近。同时用生成关系导出相似关系,并把图像分割问题转化为求解像素相似关系的等价类问题。在实现中通过求解相似关系关系图的连通分量来求解相似关系的等价类。实验结果表明本文算法可较有效地分割图像,为基于分形编码的图像分割体系提供了新的方法。

英文摘要：

An image segmentation algorithm based on fractal encoding is proposed. The working principle of the new algorithm is entirely different from existing image segmentation algorithms based on fractal encoding. The self-similarity between images blocks on different scales is utilized in the new algorithm. The generation relation of pixels is defined according the result of fractal encoding, and the error measurement is carefully selected to make the gray value of the pixels satisfying the generation relation enough similar. At the same time, the similarity relation is derived by the generation relation and the image segmentation is translated into finding the equivalent class of the similarity relation, which is equivalent to finding the connected components of the relation graph. The experiment results indicate that the new algorithm can segment images effectively and it provides a new method within the framework of image segmentations based on fractal encoding.