Abstract: In order to reduce the influence of the high linear correlation of each color component in the ＲGB( Ｒed Green Blue) space and the scale correlation of the Euclidean distance to the image segmentation results in classical k-means algorithm，the Lab color space can overcome the defect of uneven color distribution in ＲGB space，an improved k-means clustering method for color image segmentation based on Lab color space is proposed． Firstly，the color space is converted from ＲGB to Lab，and each pixel can be represented by the combination of L，a and b． Secondly，mahalanobis distance is used to replace Euclidean distance，and the improved k-means algorithm is used to cluster the pixels of the image to achieve the purpose of segmentation．Finally，experiment results show that the improved algorithm has better segmentation effect and accuracy than the classical k-means algorithm．

Key words: cluster, image segmentation, color space, mahalanobis distance