An Integrated Approach to Determination of Quality Metric for No-Reference Images

Digital images are subjected to loss of information, various kinds of distortions while processing or transmitting, which deteriorate visual quality of the images. Quality of an image plays fundamental role to take vital decision and therefore, its assessment isessential prior to application. Despite rapid advancement in technology, thecharacteristics of human vision are still considered best performer for quality assessment of images. Modelling physiological and psycho visual features of the human visual system (HVS) are reported for developing image quality assessment(IQA) methods. However, due to limited knowledge of HVS, computational HVS modelling used in IQA is far from complete. Till now the most reliable means ofassessing image quality is subjective evaluation based on the opinion of human observers. However, subjective testing is not automatic, a lengthy process and expensive too. Fuzzy based approaches have long been used to model human perception about the given tasks by transforming human observations into mathematical understanding. In the work, type-1 fuzzy inference system and fuzzy relational classifier have been built considering human perceptual vagueness in assessing quality of images using linguistic values.