Abstract: A Generalized Discriminant Analysis (GDA) based method is proposed to find the optimal chroma-like channel for image splicing detection. The channel design scheme is modeled as an optimization problem in which the objective function is the discriminative power of GDA with constraint of coefficients of the chroma-like channel. The optimization problem is resolved by grid searching combined with the gradient ascent algorithm. The results obtained from the experiment in the Columbia Image Splicing Detection Evaluation Dataset show that the detection performance of four current methods in the mainstream with regard to the rate of identification in the most optimal color designed channel is better than that in the existing color channels, which verify the versatility and effectiveness of the proposed method.