Direction based Graphical Model for Face Sketch Synthesis

Face sketch synthesis has important practical applications. Although existing face sketch synthesis methods have made great progress, these approaches ignore the direction of sketch textures. Particularly, they simply apply Euclidean distance of intensities to measure the similarity of image patches. This will result in the inconsistency of texture directions between neighboring image patches. In this paper, we propose a direction based graphical model (DGM) to improve this problem. Firstly, Gabor function is employed to extract the directions of image patches. Secondly, Markov network which combines image directions with image intensities is built to infer the synthesized sketch. Experimental results demonstrate that the proposed method achieves superior performance compared with state-of-the-art methods.