An analysis of interpolation methods for super resolution images

摘要

The image such as CT scan, x - ray image, CCTV videos and hand phone's camera is kind of low resolution image producers. Digital camera captured the continuous scenes and transform into discrete presentation in term of space and intensity. In sampling process it may create aliasing and information lost at frequency below the Nyquist sampling rates. Therefore the image suffered with an ill-posed problem by aliasing and loss of frequency. The problem ill-pose problem could be solved by applying Super Resolution (SR) techniques. The SR process contains of image registration, interpolation and image reconstruction. However this paper is focus on an analysis the best performance offered by interpolation techniques. An analysis procedure requires interpolation kernel inspection into frequency domain plotting to determine the best kernel response in pass and stop band. Otherwise use Peak Signal to Noise Ratio as indicator the similarity simulated with original image. In this study found the cubic spline interpolation is provided the smoother function frequency response with less ripples in stop band and good pass response. Besides that, it shows a superiorly in lead the highest PSNR for all type image tests with several of upscale. The best response and less distortion effect generated by kernel is preferable candidate to produce an efficient image application with low maintenances.