Abstract:
This paper addresses the problem of super resolution - obtaining
a single high-resolution image given a set of low resolution images
which are related by small displacements. We employ a reconstruction
based approach using MRF-MAP formalism, and use approximate optimization
using graph cuts to carry out the reconstruction. We also use
the same formalism to investigate high resolution expansions from single
images by deconvolution assuming that the point spread function
is known. We present a method for the estimation of the point spread
function for a given camera. Our results demonstrate that it is possible
to obtain super-resolution preserving high frequency details well beyond
the predicted limits of magnification.
Result-1: Effect of noise - The affine tranformed images are generated from the template and noise of SNR=2 is added to each. The 24 images are registered and SR algorithm is applied on them. Each input image is of size 160x128 and SR image is of 640x512 (4 times magnification in each direction)
Ground Truth
Linear Interpoltaed Reference noisy image
SR Image using Single Image Expansion Algorithm
SR Image(640x512) with 24 noisy images (160x128)
Result-2: Natural Images - The images of leaves are taken from olympus camera with translation and rotation of camera. The 16 images are registered and SR algorithm is applied on them. Each input image is of size 100x50 and SR image is of 400x200 with four times magnification in each direction
Interpolated Image
Single Image Expansion
SR Image using IBP method
SR Image (400x200) with proposed algorithm
Result-3: Text image - The images are taken from olympus camera with translation and rotation of camera. The 16 images are registered and SR algorithm is applied on them. Part of 640x560 image (550x180)
Interpolated Image
Single Image Expansion
SR Image using IBP method
SR Image with proposed algorithm
Result-4: Chart image - The images are taken from olympus camera with translation and rotation of camera. The 16 images are registered and SR algorithm is applied on them. Each LR image is of size 72x32 and SR image is 288x128
Interpolated Image
Single Image Expansion
SR Image using IBP method
SR Image with proposed algorithm