针对现有人脸画像合成算法复杂度高、合成速度慢的问题,提出了一种基于岭回归与最近邻搜索的快速人脸画像合成算法。该算法的核心是利用岭回归由人脸照片块获得初始合成画像块,并在此基础上利用最近邻搜索在训练画像块样本集中找到与初始合成画像块最相似的画像块作为最终合成画像块,使得细节纹理更具画像风格,并将合成画像块拼接得到合成人脸画像。通过仿真实验表明,该算法不仅能由人脸照片合成出高质量的人脸画像,且具有计算复杂度低、合成速度快的优点。
A fast algorithm of face sketch synthesis based on ridge regression and nearest neighbor search is proposed in this paper to address the high complexity and slow speed of existing algorithms. The initial synthesized sketch patch is obtained firstly from photo patch using ridge regression. Then the nearest sketch patch is found as the final high quality synthesized sketch from training sketch patch samples using nearest neighbor search. The experimental results show that the proposed algorithm offers synthesis of high quality face sketches with low complexity and at high speed.