Learning a blind quality evaluation engine of screen content images

Abstract

a b s t r a c tWe in this paper investigate how to blindly predict the visual quality of a screen content image (SCI). With the popularity of multi-client and remote-controlling systems, SCIs and the relevant applications have been a hot research topic. In general, SCIs contain texts or graphics in cartoons, ebooks or captures of computer screens. As for blind quality assessment (QA) of natural scene images (NSIs), it has been well established since NSIs possess certain statistical properties. SCIs however do not have reliable statistic models so far and thus the associated blind QA task is hard to be addressed. Aiming at solving this problem, we first extract 13 perceptual-inspired features with the free energy based brain theory and structural degradation model. In order to avoid the overfitting and guarantee the independence of training and testing samples, we then collect 100,000 images and use their objective quality scores computed via a high-accuracy full-reference QA method for SCIs as labels, before learning a new blind quality measure from aforementioned 13 features to the objective quality score. Experimental results performed on a large-scale screen image quality assessment database (SIQAD) demonstrate that the proposed blind quality metric has a good correlation with human perception of quality, even superior to state-of-the-art full-, reduced-and no-reference QA algorithms.

Learning a No-Reference Quality Assessment Model of Enhanced Images With Big Data

“…Recently, a new strategy has been proposed towards finding the regression module in blind IQA designs [55]. To be more specific, in order to overcome the issue of overfitting, greater than 100,000 images are utilized as training samples to learn the regression module in our blind BIQME metric.…”

“…How to label these generated images? In [55], Gu et al indicated that, rather than training on human opinion ratings, using predicted scores computed from high-performance FR-IQA methods as training labels is a good choice. The lately proposed PCQI metric was proven to highly correlate with subjective quality scores on enhancement-relevant databases [35], but it does not take the influence of colorfulness into consideration, which is obviously an important index of image quality.…”

“…The second type consists of two RR-IQA models, RRED [15] and FTQM [16]. The third type contains BRISQUE [17], NFERM [18], NIQE [19], IL-NIQE [20] and BQMS [55] without access to original references in assessing the visual quality of images. The last one consists of FR C-PCQI, RR RIQMC [30], RR QMC [34], blind FANG [36], and blind GISTCM [37], which are dedicated to enhanced IQA tasks.…”

Quality assessment metric of stereo images considering cyclopean integration and visual saliency

“…Furthermore, enhancement technologies may possibly play a more significant role in future IQA research, since they are able to generate betterlooking images, even outperforming the natural images that are deemed to have the optimal quality. At the same time, further research is needed in screen content images, which are becoming particularly important due to their applications in remote computing and cloud gaming [19,14,55].…”

“…This means that the weighting cyclopean saliency maps will be replaced by the saliency map derived from either view (left or right) only in equation (18) and (19). Fig.…”