FISBLIM: A FIve-Step BLInd Metric for quality assessment of multiply distorted images

The last decade has seen a surge of interest in the research of image quality assessment (IQA). Many successful quality metrics, such as structural similarity index (SSIM) are reportedly to achieve very high accuracy for various kinds of image distortions. However, in practice, multiple image distortions tend to occur together and this leads difficulty to previous works of IQA including SSIM and variations. This problem is even more difficult for no-reference or blind quality assessment. To answer this challenge, this paper proposes a new FIve-Step BLInd Metric (FISBLIM) for quality assessment of multiply distorted images. The algorithm is built upon several common image processing blocks to simulate the image perceiving process of the human visual system (HVS). The FISBLIM method is not training based and the performance is robust and not database-dependent. Experimental results on the newly released LIVE multiply distorted image quality database demonstrate the effectiveness of FISBLIM as compared with mainstream full-reference and no-reference image quality metrics.