Investigating a relationship for MISO reconstruction-based image super resolution techniques using limited data

This paper investigate a relationship existed between the number of low resolution input images and the resultant high resolution image for existing multi-input single-output (MISO) reconstruction-based image super resolution techniques. Ideally, the quality of the reconstructed high resolution image is directly proportional to the number of low resolution input images. However in practical scenario, the number of low resolution input images should be kept as low as possible to avoid the error accumulation in the reconstruction process. Two new concepts, the point of saturation (PoS) and point of cut off (PoC) for super resolution techniques have been introduce in this paper. The point of saturation occur because all the input images have not addition information for the reconstruction process. One of the main reason for point of cut off is mis-registration. The existence of both those points vary from method to method as each technique deal the reconstruction process in different way. Both of them are shown on a performance graph in order to judge the performance of existing super resolution algorithms. Our study were limited to small number of inputs as we considered it more practical situation. From our Experiments, we observed that PoC occur more times than PoS.