一种基于边缘检测的局部阈值分割算法

本文提出了一种基于边缘检测的局部阈值分割方法.该方法将整幅灰度图像分成小块,在每个小块中利用梯度算子对小块中的边界点进行检测,寻找出小块内的所有边界点;然后沿着这些边界点的梯度方向找出最临近的点,以所有这些临近点和边界点的灰度均值作为该小块的灰度阈值进行分割.该算法计算复杂度较低,避开了灰度直方图阚值分割方法中“谷底”难以确定的问题,同时照顾到了图像的局部灰度特性.

A Local-Threshold Segment Algorithm Based on Edge-detection

In this paper we present a multi-threshold segment algorithm based on Edge-detection. According to this method, we divide a gray-level image into many small blocks, and find out all may-be edge pixels using gradient operator in each block. For each edge pixel P, we try to find another nearest pixel Q along the gradient-direction. We use the average of edge pixel P and near pixel Q as the local threshold in the respective small block. This algorithm has lower computing-complexity, bypasses the issue that it is not easy to detect the proper "valley" in gray-level histogram of image and also overcomes the shorting coming of using only one global threshold.