基于度信息的图像过渡区提取与分割

Degree-based image transition region extraction and its segmentation

提出一种基于度信息的图像过渡区提取与分割算法 .将一幅图像映射为带权无向图 ,权函数定义为节点之间的相似度 ,由于目标内部与背景内部的像素同质性好 ,相似度大 ,因此度值较大 ,而边缘像素的度值就小 ,由此提取过渡区并进行图像分割 .实验结果表明 ,本方法能够稳定地提取过渡区 。

A degree-based transition region extraction method (D-TREM) was proposed. If an image was converted into a weighted undirected graph, and the similarity between nodes the weight function defined, the degree of pixels inside the object and the background may be larger than those within the transition region. These properties of degree lead to our D-TREM. The experimental results demonstrated that D-TREM could steadily extract the transition region and the speed was significantly faster compared with other TRE method.