Image Super Resolution by Fast Edge Adaptive Interpolation

Abstract

iii Abstract Super-resolution is a term given to a set of techniques which process and enhance the resolution of images. Thus it is the process of generating an image with a higher resolution than the original input image, by estimating the pixel intensities on an up-sampled grid by various means. In this dissertation, Introduction gives a clear idea about the image processing operations known to be applied in various applications dealing with images, along with the brief introduction about image super-resolution by interpolation techniques. Literature survey provides an insight on the previous related work that has been done in the field of image zooming interpolation techniques. Lot of side effects appear when an image’s resolution is increased, in order to reduce those artifacts and produce crisp images a new algorithm is proposed which is based on the combination of three different techniques. The complete set of steps involved in the method results in super resolved images with a comparable appearance like that of original, with much clear, detailed view and least artifacts which transpire with previous methods. The colors in the various patterns are not mixed into the adjoining areas with contrasting colors, thus reducing edge blurring and providing a smooth and presentable natural-image. The improvement provided by the proposed algorithm over the existing algorithms is evidently visible in the HR images and also through a quantitative measure known as Peak Signal to Noise Ratio. The proposed algorithm provides maximum PSNR improvement of 6.53 db over MEDI, 7.80 dB over NEDI, 6.26 dB over SAI and 5.41 dB over iNEDI. This implementation is well suited for applications that need low turn-around time.