Complexity reduction in multi-dictionary based single-image superresolution reconstruction via pahse congtuency

Compared with single dictionary, multi-dictionary method can achieve better performance in image superresolution reconstruction (SR). However, the computational cost of multi-dictionary based SR is very heavy and usually time-consuming and resource-intensive. In this paper, we proposed a complexity reduction method in multi-dictionary based SR via phase congruency. The PC map of the LR image is extracted and binarized to distinct the importance of the image patches of it. Then, the corresponding important HR patches are reconstructed by multi-dictionary based SR method and the unimportant ones by single-dictionary based SR. The finalized reconstructed HR image is obtained by averaging the overlapped region between the adjacent patches. Experimental results show that our method can not only obtain competitive results but also can save much time and reduce the computational complexity in the reconstruction process compared with multi-dictionary based SR method.