An effective framework of IBP for single facial image super resolution

Super-resolution was used to refine the LRI (low resolution image) which was acquired fro m the existing electric devices. Moreover, it is significant to do this work on the human face. To solve the special facial problem, a novel framework based on IBP (Iterative Back Projection) is proposed, in wh ich the edge detection and edge preserving are taken into account. In the iterative part, we adopt wavelet for edge detection to preserve the details and eliminate the loss of visual informat ion. Furthermo re, PCA (Principle Co mponent Analysis) technique is introduced to extract more local high frequency informat ion to refine the LRI . Experimental results show that our framework could efficiently enhance the visual effect. Both PSNR and SSIM of the super-resolution of LRI could be increased. Moreover, some experiments on face recognition are conducted and our method achieves more inspiring results.