Super Resolution Image Reconstruction By Adaptive Auto Regressive Model
Katragadda Vamsi Krishna, Barjinder Singh Saini · 2015
In this paper, an Adaptive Auto regressive Model is proposed for the super resolution image reconstruction problem in which parameters of the model vary according to image. Firstly estimation of texture correlation in the image is made by measuring the co-occurrence of patterns in image and then based on the image content determination of the patch size for algorithm is done which efficiently estimates the missing pixels of interpolation grid. This method performs very well as we try to estimate the missing pixel of image by considering repetitive patterns & the nonlocal pixels patch estimations in the image retrieval process. The extent of nonlocal patch inclusion is determined by the co-occurrence in the image. This method interpolates the image by varying patch size adaptively for each image. Maximum extent of similarities in an image was taken into account for making the sparse based image interpolation technique into adaptive method. Measuring Perceptual quality metric Peak Signal to Noise Ratio (PSNR) has shown maximum increment of 0.43 db and average of 0.2 db on our experimental results proved it as best interpolation technique.