基于引导图像滤波的素描人脸图像合成技术

Sketch facial image synthesis technology based on guided image filtering

针对传统人脸图像合成算法存在过度平滑的问题,文中提出一种使用引导图像滤波的改进合成算法。通过引入K近邻基准算法和机器学习等理论,测试人脸图像可以得到训练图像块的线性组合表示,从而获取初始的合成结果。并在此基础上,利用引导图像滤波的方法,增强人脸素描图像的细节特征,进一步优化人脸细节纹理的展示。相关的仿真结果表明,文中所提出的合成算法有效回避了传统合成算法过度平滑的弊端,能够合成效果更加优秀的人脸素描图像。

In allusion to the excess smoothness in the traditional facial image synthesis algorithm,an improved synthesis algorithm using guided image filtering is proposed.By introducing the k⁃nearest⁃neighbor benchmark algorithm and machine learning theory,the face image is tested to get the linear combination representation of the training image block,so as to obtain the initial synthesis results.On this basis,the detail features of the face sketch image is enhanced by means of the method of guided image filtering,and the display of the face detail texture is further optimized.The simulation results show that the proposed synthesis algorithm can effectively improve the excess smoothness of the traditional synthesis algorithm,and can synthesize excellent face sketch images.