基于支持向量机的多特征彩色图像边缘检测

摘要

Color image edge detection in special fields requires not only accurately finding out the target edge, but also removing the non-target edge. So multi-feature color image edge detection algorithm for support vector machine is proposed, which forms feature vector with multi-feature values, combining with weighted gradient value and edge pixel neighborhood information in image chrominance and luminance channels, then applying the trained support vector machine for recognizing the target edge. Experiments show that this method has better ability to identify target edge than the traditional edge detection method.%针对在特殊领域中彩色图像边缘检测,不仅需要准确地检测到目标边缘而且需要去除非目标边缘,提出了一种新的支持向量机多特征彩色图像边缘检测方法.这种方法根据彩色图像边缘的特点,在图像亮度和色度通道上结合像素加权梯度值和像素邻域相关信息构建多维特征向量,通过训练的支持向量机可以准确识别出目标边缘.实验结果表明,该方法比传统边缘检测方法具有更好目标边缘识别能力.