Improving Robustness of Image Quality Measurement with Degradation Classification and Machine Learning

Abstract

Abstract-Image quality metrics can be classified as generic or degradation specific. Degradation specific measures perform poorly under "mismatched" conditions. Generic measures, on the other hand, may compromise quality measurement accuracy while gaining robustness to variation in distortion conditions. To improve the accuracy-robustness tradeoff, we employ supportvector degradation classification and machine learning tools to judiciously combine generic and degradation specific measures. To test our algorithm, composite quality metrics are optimized for five different distortion classes. Experiment results show that the proposed algorithm achieves improved performance and robustness relative to two benchmark generic quality metrics.

A ParaBoost stereoscopic image quality assessment (PBSIQA) system

“…Similarity) [37] HVS HVS-MSE [34] NCC (Normalized Cross Correlation) [37] View synthesis NDSE (Noticeable Depth Synthesis Error) [43] distortion levels ([1, 1], [2,2], [3,3]) and three asymmetric distortion levels ( [1,2], [1,3], [2,3]). We conducted a subjective test to obtain MOS using the absolute category rating (ACR) [28].…”

“…Based on previous studies on image quality assessment [34], [35], [37], [38], [39], [40] and our own experience, we select 24 candidate features for further examination as listed in Table II. Then, we calculated the Pearson correlation coefficient (PCC) to indicate the prediction performance between MOS and a single-feature-based quality scorer (with the exception that the singular value and the singular vector are integrated into one feature vector for the SVD scorer) over Datasets A and B, respectively.…”

“…Such artifacts are called the information additive distortion (IAD) [20]. As stated in [10], [34], there are two good features in detecting blockiness. They are the HVS-modified MSE (HVS MSE) and the zero crossing rate (ZCR).…”

“…They are the HVS-modified MSE (HVS MSE) and the zero crossing rate (ZCR). Mathematically, the HVS MSE [34] is given by…”

Deep multi-task learning for image/video distortions identification

“…It is therefore very interesting to develop an automatic method that mimic this HVS's capability. In recent years, much efforts have been devoted to the problem of identifying distortion in images and videos [18][19][20][21][22][23][24][25][26][27][28][29][30][31][32][33][34]. Most studies have focused on identifying a single type of distortion [26], however, images and videos can suffer from multiple distortions in most real-world applications [38], which is more challenging to address given the complex interactions and masking effect among distortions.…”