摘要： 针对图像分割过程中前背景表面特征不均质问题，提出一种基于颜色纹理先验特征的多通道局部能量模型。对现有局部能量模型进行扩展，降低初始轮廓线位置对分割结果的影响，并引入8维HSV颜色模型和变换域结构张量纹理特征，实现前背景颜色特性相似的图像分割。实验结果表明，该模型具有较好的分割效果。

关键词: 图像分割, 主动轮廓模型, 局部能量, 水平集, 轮廓波变换, 结构张量

Abstract: In this paper, a new localized energy active contour model combined with multiple features is proposed for image segmentation, especially for those images with the inhomogenous appearance features in foreground and background. It reformulates the localized segmentation energy to enhance its robustness to initialization. By introducing more fine 8-d HSV color space and structure tensor texture features of transform domain, it is able to segment images with weak object boundaries, images with visually similar color properties in foreground and background. Experimental result shows that the proposed model is exemplified on various images objects and its superiority over state of the art variational segmentation techniques is demonstrated.

Key words: image segmentation, active contour model, localized energy, level set, contourlet transform, structure tensor

中图分类号: 

• TP911.73