Performance Analysis of Wavelet Transforms for Learning based Single frame Image Super-resolution

super resolution concept has been introduced for image enhancement in various applications. Image enhancement is crucial operation essential for reducing different possible degradations of the captured image. More sophisticated techniques are already proposed. Wavelet transform based algorithms are widely used in many applications. Wavelet transform/s is used to extrapolate missing high frequency components which improve the efficiency of an algorithm. In this paper for super resolving the images, wavelet coefficients of the unknown high resolution image are learnt from a set of high resolution training images in wavelet domain. The performance of different discrete orthogonal and a biorthogonal wavelets have been evaluated on different class of images in terms of MSE and PSNR. The outcome of this work suggests that use of db4 wavelet transform is appropriate for super resolution technique. The PSNR obtained with this transform outfits for other wavelet transforms.