Abstract: In order to obtain high quality artistic Mosaic collage visual effect, a parallel collage algorithm based on histogram matching was proposed. According to the histogram distance, the best matching sub-image of the original image segmentation block was selected, then the histogram matching method was used for color correction. In the collage algorithm, segmentation of original images, comparison of histograms, color correction and image filling were all parallelly performed with GPU (graphic processing unit) based on compute unified device architecture (CUDA). The effectiveness of the proposed algorithm was verified by experimental analysis and user survey. The results show that compared with the traditional color correction method based on color mean, the collage result of this algorithm is closer to the original image, and the overall visual effect is more coordinated. GPU parallel computing based on CUDA improves the efficiency of the algorithm, thus realizing the recursive collage of interactive art Mosaic images.