Reduced-Reference Image Quality Assessment With Visual Information Fidelity

Abstract

Reduced-reference (RR) image quality assessment (IQA) method aims to accurately measure quality with part of the reference data. The challenge for RR IQA is how to effectively represent the visual content of an image with limited data for quality measurement. Inspired by the orientation selectivity (OS) mechanism in the primary visual cortex, we introduce an OS based visual pattern (OSVP) to extract visual content for RR IQA. The OS arises from the arrangement of the excitatory and inhibitory interactions among connected cortical neurons. Inspired by this, we investigate the correlation among neighbor pixels, and propose the OSVP to represent the visual content of an image. Then, the quality degradation is measured as the changes of OSVP, and a novel RR IQA model is proposed. Experimental results demonstrate that the OSVP based RR IQA model uses limited reference data (9 values) and performs highly consistent with the subjective perception.

Perceptual Quality Measure of Screen Content Images

“…By adopting subjective testing strategies ,several image quality assessment databases [8]- [9] have been constructed, various Full Reference (FR) IQA methods [10], [11]- [12], such as SSIM, VIF, FSIM, MAD, GSIM and GMSD, have been proposed to objectively assess the quality of distorted natural images. Besides, Reduced Reference (RR) IQA [13] and No Reference (NR) IQA metrics [14] are also reported. Document Image Quality Assessment (DIQA) has got attention in the research community recently due to the increasing requirements of digitization of historical or other typewritten documents.…”

Blind Image Quality Assessment on Gaussian Blur Images

“…To this end, various IQA methods have been proposed. According to the usage of the reference image, objective IQA metrics can be divided into three categories: full reference (FR) [4,5], reduced-reference (RR) [6], and blind/no-reference (NR) [7]. amplitudes and different frequency signals.…”

Blind image quality assessment with improved natural scene statistics model

“…Recently, full-reference (FR) image quality assessment (IQA) metrics have been demonstrated to perform highly consistent with human perception [3][4][5][6]. Reduced-reference (RR) IQA metrics require partial information from the reference images passed through an ancillary channel and have achieved great progress [7][8][9][10]. FR and RR metrics assume that the whole or partial reference information is known beforehand [11].…”