The full reference quality assessment metrics for super resolution of an image: Shedding light or casting shadows?

There is a driving need for digital images of high resolution and quality in different Multimedia applications. However, a high resolution digital camera is still very expensive in the market. To meet such demand of today, the super resolution techniques promises to provide a high resolution image for multiple low resolution input images. But to judge the performance of existing super resolution algorithms objectively is a difficult task because lack of sufficient realiable testing tools. In this paper, it is shown that the existing Full Reference (FR) metrics cast shadows on the overall quality of super resolved image so judging super resolution quality using them is not an good idea. The main reasons for that is they do not take the visual masking phenomenon into consideration and also, they depends on the reference input, which itself may be of low quality. Experimental results shows that there is a need for a new Full Reference metric that shed light on the quality of reconstructed image in order to compare the performance of different existing super resolution (SR) methods.