A low computation complexity, which is a very efficient representation of image for visual classification tasks, is presented. The collaborative representation was combined with discriminative ingredient in a unified framework, which is an extended version of collaborative representation-based classification. The coefficients of collaborative representation of test samples are sparse and robust to occlusion or other disguises based on redundant and over-complete dictionary. Besides, the discriminative information was exploited by minimizing the within-class scatter and maximizing the between-class scatter, which is very helpful for visual classification tasks. Experimental results on some widely used benchmark datasets indicate that the proposed method can achieve competitive performance with other existing works.