A no-reference computer-generated images quality metric and its application to denoising

A no-reference image quality metric detecting both blur and noise is proposed in this paper. The proposed metric is based on IFS2 entropy applied on computer-generated images and does not require any edge detection. Its value drops either when the test image becomes blurred or corrupted by random noise. It can be thought of as an indicator of the signal to noise ratio of the image. Experiments using synthetic, natural and computer-generated images are presented to demonstrate the effectiveness and robustness of this metric. The proposed measure has been too compared with full-reference quality measures (or faithfullness measures) like SSIM and gives satisfactory performance.

I. Introduction

No-reference image quality measures play a significant role in image processing (video compression, [1], [2], algorithm performance analysis, algorithm parameters optimiza-tion, communication, …). These measures are also of crucial importance in computer-generated images, where they are still in investigation. Some recent papers [1], [2] proposed noreference quality assessment of JPEG images. Although the authors obtained good results, these reported quality measures have their limitations. At the moment, the classical model to characterize image quality remains psycho-visual experiments (Human in the loop experiment [3]).