Blind super-resolution image reconstruction based on novel blur type identification

Blind super-resolution image reconstruction is to obtain a high-resolution image from a sequence of low-resolution images which are degraded by unknown blur, noise, and down sample. Conventional super-resolution image reconstruction algorithms assumed that the blur type is known, thus automatic blur identification is of important significance in blind superresolution image reconstruction. This paper proposed a novel blur type identification algorithm for blind image superresolution. The proposed blur type identification method uses a dictionary learning to identify three blur kernels. It includes the logarithmic normalized feature matrix, the structural similarity index, and the best structural similarity between observed images and dictionary images. Furthermore, we applied the proposed blur type identification method to blind image super-resolution. The experimental result shows that the identification accuracy of proposed method can achieve 98% above. More importantly, the proposed blur type identification-based algorithm for blind image super-resolution can enhance the performance of reconstruction quality according to visual quality and evaluation index.