| 摘要 |
A method for automatically denoising a two-dimensional color face image using recursive PCA(Principal Component Analysis) reconstruction is provided to remove pimples, scars, freckles and skin troubles from a face while maintaining structural characteristic of the face. A face space is learnt using PCA. Automatic face feature extraction using a dynamic appearance model is performed. A reconstructed color image is re-lighted using a bilateral filter. A noise region is extracted using variances of training data. An image is reconstructed using information on a part of an input image, and the reconstructed image is blended with an original image to generate a noise-free image. |
