A Method of Character Expression Animation Deburring Based on PCNN
The method uses PCNN network to segment the captured video frame images,remove the background part of the image,and input the retained part of the face expression into the improved deep learning model to extract the face expression features.Bidirectional constraints of style and content are intro-duced to determine the model loss function.All the extracted expression features are fused to generate the character expression animation.The region reconstruction algorithm is used to remove the edge burrs of the generated animation to ensure the animation generation effect.The test results show that:the image segmen-tation performance is good,and the results of the intra-region uniformity criterion are all above 0.924,which can clearly retain part of the face expression image and generate expression animation of different characters based on the real face expression;and the maximum values of the shape measurement criterion and the Pearson's correlation coefficient indicator are 0.994 and 0.957 respectively,which can effectively remove the edge burr of the animation to ensure the quality of the animation.
character expression animationdeburringgeneration methodfeature fusionimage seg-mentationbidirectional constraint
万方数据
维普
