Multi-frame super-resolution using adaptive normalized convolution

An enhanced fusion algorithm for generating a super-resolution image from a sequence of low-resolution images captured from identical scene apparently a video based on adaptive normalized convolution has been designed and analyzed. The algorithm for fusing the images is based on the supporting structure of normalized convolution. Here the idea is projection of local signals onto a subspace. The adaptive nature of the window function in adaptive normalized convolution helps to gather more samples for processing and increases signal-to-noise ratio, decreases diffusion through discontinuities. The validation of proposed method is done using simulated experiments and real-time experiments. These experimental results are compared with various latest techniques using performance measures like peak signal-to-noise ratio, sharpness index and blind image quality index. In both the cases of experiments, the proposed adaptive normalized convolution-based super-resolution image reconstruction has proved to be highly efficient which is needed for satellite imaging, medical imaging diagnosis, military surveillance, remote sensing, etc.