Based on basic properties of L*A*B* color space, this paper proposes a new approach of 2D image processing which is used for measurement of injury rate on fish by a modified K-means clustering algorithm and Otsu’s threshold algorithm. Then, experimental results of the proposed method are compared to the results of a manual threshold method on L*A*B* color space. To do this issue, the following tasks are done. Firstly, an original color image is transferred into L*A*B color space. Secondly, channel “a” is separated from L*A*B* color space image. Thirdly, the value of channel “a” is adjusted by changing the contrast algorithm. Fourthly, the modified K-means clustering algorithm on a new channel “a” image is applied to define and divide data elements into different groups. Fifthly, Gaussian Filter is used to filter the random “noises” in shape of injury and fish images. Sixthly, Otsu’s threshold algorithm is used to transfers the filtered images into binary images. Seventhly, final images are obtained after filtering the rest of “noises” by morphological processing. Finally, the areas of injury and fish shapes are obtained by counting pixels on both of the final binary images. The experiment results show that the proposed new approach is closer to the real injury and injury rate on fish than the results of the manual threshold method on L*A*B color image.