Specific Wavelet Family Selection for Wavelet Domain-Based Super-Resolution Application

Abstract

The prime need of image quality for accurate analysis with precise decisions is the key attraction for researchers in image processing field. The increasing need is due to degradation of image signals whilst capturing, transmission, compression, etc. The accuracy in super-resolution process for quality improvement of images or videos is achieved at the cost of time and complexity. This limitation is motivation for researcher to contribute themselves in the same field by developing conventional algorithms and methods specially categorised in spatial and frequency domain. In the recent decades, wavelet domain processing has remarkable results in super-resolution field. The results of wavelet domain processing are depending upon wavelet functions or families considered whilst analysis, as these families possess unique properties which gives different results according to application area. Unfortunately, there is no such theory or analysis available which shows specific wavelet family selection for particular super-resolution process. The author has tried to explore the concept behind selection of appropriate wavelet function with analysis on different video frames containing variety of scenes. The process is simple, i.e. decomposition and reconstruction of video frames using different wavelet families and comparative analysis of original and reconstructed image/frames with different quality measurement metrices. The exact reconstruction shows lossless wavelet domain process. The winner wavelet function is Haar which is simplest amongst all wavelets, despite the literature provided in wavelet processing preferred db2/db7/9 families. The assessment provided is beneficial for beginners to select appropriate wavelet function in super-resolution application.