【摘要】 目的基于子波变换研究纹理图像的多分辨率分割．方法提出了子波域主能量通道的纹理图像多级分割方法．以最大能量为线索，建立起由细到粗的子波分解金字塔，分割按照其逆过程，即由粗到细，将低分辨率上的分割结果传递到高分辨率上．结果在每个分辨率上优化选择分割特征，用模糊Ｃ聚类法确定分割方案，有利于纹理图像的分割，分割结果逐级精细．结论员分辨率上基本无纹理栅格，分割效果符合人眼视觉特性．更多还原

【Abstract】 Aim Multi-resolution segmentation using wavelet transform is studied. Meth- ods A hierarchical segmentation based on the subband image with the most energy in wavelet domain is suggested.The scheme includes the construction of a fine to coarse multi-resolution wavelet pyramid by energy, and then the segmentation of textural image by the inverse process of wavelet decomposition, i. e., the process from coarse to fine resolution. The segmentation information at a lower level is propagated to the next higher level. Results Feature optimal selection and the use of Fuz- zy-C clustering are helpful to segmentation at every resolution. The result gets finer with the levels. Conclusion There are almost no texturel grids at the higher levels, so the segmentation accords well with human vision.更多还原