一种新的自动多阈值图像分割方法

A New Image Segmentation Approach by Using Automatic Multi-threshold Selecton

本文提出了一种新的自动阈值图像分割方法。该方法选择二维灰度直方图的最佳一维投影作为输入信息，并利用了小波变换的多尺度特征分析此投影．在大尺度下抑制噪声并自动确定分割类数，在小尺度下精确定位阈值。实验表明新方法能够自动而快速地对图像进行分割而且有较强的抗干扰能力．

This paper presents a new image segmentaton approach by suing automatic multi-threshold Selection. the approach uses the 1-dimension optimal projection of 2-bimension gray histogram as the inputinformation and the wavelets to analyze this projection. At a large and, the noise is restricted and the number ofthresholds is automatically computed. At a small ot, the precise positions of thresholds are decided. Experimental results show this new approach can segment images automatically and quickly.