基于神经网络的灰度图像阈值分割方法

A Thresholding Method for Grey Image Segmentation Based on Neural Network

对于一幅目标区域和背景区域在灰度上有较明显差异的图像,为了把目标从背景中分离出来,常利用直方图谷点作为分割阈值来分割图像.本文采用一种基于神经网络的线性搜索的方法来确定直方图的谷值可以得到最优的阈值来分割图像,克服了用常规的极小值阈值法确定的极小值不稳定和不可靠的缺点,并对实际图像进行分割处理.实验结果表明采用这种新方法来搜索谷值可以得到全局极小值,用于分割图像可以把目标从背景中分离出来,并取得了满意的结果.

As for an image with different gray values in the object and the background, in order to separate the object from the background, the minimum of the histogram is often used as the threshold to separate the image. In this paper, a new method is adopted based on neural network＇ s linear searching to determine the minimum of histogram. Through this method the optimal threshold is obtained and the instability and the unreliability of the minimum are conquered, which are determined by the thresholding metheod of the minimum. Experimental results show that adopting this new method to search for the minimum can obtain the absolute minimum, and using the optimal value to segment the image can separate the object from the background and get the satisfactory results.