基于特征滤波的生成对抗网络素描人脸合成方法

SKETCH FACE SYNTHESIS METHOD BASED ON FEATURE FILTERING GENERATIVE ADVERSARIAL NETWORKS

  • 摘要: 素描人脸合成技术在各领域发挥着重要的作用。传统素描人脸合成方法生成的人脸轮廓清晰度不足,纹理模糊,轮廓内的面部特征细节有所缺失,同时,图像中还存在着明显的粗糙像素点。为了解决上述问题,在生成对抗网络的基础上提出了一种基于特征滤波的网络模型。该模型首先通过提取多层级面部特征丰富了人脸特征细节,然后通过外观滤波与纹理处理模块改善了人脸的外观轮廓与纹理表现,最后通过全变差损失函数减少了图像中的粗糙像素点,提高了图像整体清晰度。通过在CUFS和CUFSF人脸数据集上的实验,证明了该方法在丰富人脸特征、改善纹理细节、减少图像粗糙像素点等方面的有效性。

     

    Abstract: Sketch face synthesis technology plays an important role in different fields. The sharpness of the face contour generated by the traditional sketch face synthesis method is insufficient, the texture is blurred, and the details of face features inside the contour are missing. At the same time, there are obvious rough pixels in the image. To solve the above problems, a generative adversarial network model based on feature filtering is proposed. This model extracted the features of different levels of faces to enriches the feature details, improved the appearance contour and texture expression of faces through the appearance filtering and texture processing module, and reduced the rough pixels in the image through the total variation loss function, thus improving the overall clarity of the image. Experiments on the CUFS and CUFSF face datasets prove the effectiveness of this method in enriching face features, improving texture details, and reducing image rough pixels.

     

/

返回文章
返回