基于自组织特征映射网的彩色图像二值化方法

本文提出一种基于自组织特征映射网(SOFM)的彩色图像二值化方法。使用彩色像素的RGB值训练网络权值,找到图像的目标类RGB中心和背景类RGB中心,再利用竞争取胜的机制处理每一个像素使其二值化。为了减少算法处理的数据量,采用决策表属性简化方法对图像的颜色特征:进行简化。由于算法直接利用彩色图像的原始特征信息,二值化过程不受灰度化算法的影响。应用实例证明了本文的算法具有好的效果。

The Binary-Conversion Method of Color Image Based on SOFM

A binary-conversion method of the color image is proposed based on SOFM. It finds the RGB center of the aim class and the RGB center of the background class by training the network with the RGB values of the color image elements. Then every image element is converted into binary through the competition in SOFM. In order to simplify the data processing, the color features are reduced by the attribute reduction method of the decision table. Since the method directly utilizes the original features of the color image, the binary-conversion is uninfluenced by the gray-conversion algorithms. The application examples have shown a good result.