自适应图像边缘检测的并行算法

A PARALLEL METHOD TO DETECT BOUNDARIES IN GREY LEVEL IMAGES

本文提出一个新的图像边缘检测的并行算法。该算法的要点是根据对图像中各邻近像素对之间的亮度级差值所作的统计,得出反映这种差值分布的直方图,称为D-F直方图。以这个直方图为基础,通过最大熵(entropy)的计算,可确定能用于检测图像边缘的阈值。将这个阈值作为一个算子,并行地作用于图像的各邻近像素对上,可以提取图像上各物体的边缘。本文还提出自适应的图像分块方法,以便对复杂的图像得到较满意的边缘输出。

This edge detecting method is based on Shannon's entropy concept. We introduce a new histogram called D-F histogram and then obtain an edge detector by calculating the maximum entropy of the D-F histogram. To avoid the disadvantage of using a single threshold over the whole image, the technique is then applied to sub--images obtained by dividing the original scene into local regions. This enables us to use a variable detector over the whole scene and provides an increase in parallelism and a better result for edge detection.