关键词: 图像二值化, k中心点聚类, 阈值

Abstract: In the research on machine vision and pattern recognition, transforming the image into two-value image is the key foundation to more efficiently recognize specific area or target of image. This paper presents an image binarization processing method using k-medoids clustering. This method uses square sum error based on distance as the clustering quality metric function. According to the field knowledge of image binarization, this method sets the value of k as 2, divides the image into foreground class and background class. The experimental results show that, for natural images under complex environment, this method is better than OSTU thresholding method in the effectiveness and efficiency.

Key words: image binarization, k-medoids clustering, threshold