Abstract: The purpose of image reconstruction is to estimate a high-resolution image with better vision effect from one or a sequence of images. Super resolution that comes from previous image restoration and reconstruction technology can take advantage of redundant information among images. Super resolution reconstruction could include two kinds of technology; reconstruction-based technology and learning-based technology. The reconstruction-based technology estimates a high-resolution image from input images according to specific degradation model. Learning-based technology supplies input images with prior knowledge from training examples to get a better result. The paper systematically introduced algorithms of super resolution technology. Finally, we pointed out that image registration, reconstruction of degradation model and learning model, blind estimation, learning algorithms were main problems and future direction in image super resolution technology.

Key words: Super resolution, Degradation model, Learning model, MRF, MAP