摘要： 针对包含多个目标或目标灰度与背景灰度接近的图像分割问题，借鉴人类视觉系统的关注机制和多分辨性，提出一种多尺度框架下基于感兴趣区域提取的图像水平集分割方法。对原图像小波变换的低频分量图应用显著性特征提取出感兴趣区域，将图像域分成多个感兴趣子区域和一个背景子区域，在各目标子区域中，采用C-V模型方法进行曲线演化，并对各子区域分割结果进行合成。仿真结果标明，该算法能有效分割多目标图像。

关键词: 多目标分割, 水平集, 感兴趣区

Abstract: Aiming at the problem that the segmentation of image with multiple objects more than two different mean intensities is a nodus, an approach for image segmentation is presented based on the multi-scale analysis and extraction of Region Of Interest(ROI), which is put forward according to the attention mechanism and multiresolution of human vision system. Wavelet transform is used to get the low frequency sub-image, and low frequency domain is divided into multiple sub-regions, and then C-V model is employed for curve evolution in each sub-region. Simulation results demonstrate the feasibility for segmentation on multiple objects and low contrast image of the proposed method.

Key words: multi-objects segmentation, level set, Region Of Interest(ROI)

中图分类号: 

• TP391.41