基于侧抑制竞争原理提取图像边缘的方法

A New and Better Way of Extracting Digital Image Edge

根据动物侧抑制原理，提出数字图像侧抑制竞争提取边缘方法。首先提出数字图像侧抑制网络基本结构，确定必要约束条件和学习方法，再在图像中选定若干边缘模块作为样本输入网络学习，得到抑制竞争网络的连接权值，最后将图像输入网络，得到图像边缘。分析结果表明，这种方法具有提取速度快，边缘不移位，能用硬件并行处理实现等特点，可用于实时提取图像边缘。

Digital image edge is a principal feature of many signals, including infrared, radar, and visible - light signals. It is also strongly stable. So in target identification, we often extract digital image edge.Inhibitory interaction allows animals to see images better in several ways: (1) no distortion, (2) rapidity in seeing, (3) robustness against disturbance. We utilize the principle of inhibitory interaction to improve the extraction of digitai image edge.First, we designed a basic construction of inhibitory interaction networks. Then we determined the necessary constraints and learning strategy. After learning with several tem plate digital images, we assigned different inhibitory interaction weights to different units of the inhibitory interaction networks. The finally adjusted networks is effective in extracting digital image edges. It possesses high computing efficiency, high fidelity of image edge, and robustness against disturbance. It is capable of parallel processing. So ,it can extract digital image edges real time.