Image quality assessment based on a degradation model

Abstract

We model a degraded image as an original image that has been subject to linear frequency distortion and additive noise injection. Since the psychovisual effects of frequency distortion and noise injection are independent, we decouple these two sources of degradation and measure their effect on the human visual system. We develop a distortion measure (DM) of the effect of frequency distortion, and a noise quality measure (NQM) of the effect of additive noise. The NQM, which is based on Peli's (1990) contrast pyramid, takes into account the following: 1) variation in contrast sensitivity with distance, image dimensions, and spatial frequency; 2) variation in the local luminance mean; 3) contrast interaction between spatial frequencies; 4) contrast masking effects. For additive noise, we demonstrate that the nonlinear NQM is a better measure of visual quality than peak signal-to noise ratio (PSNR) and linear quality measures. We compute the DM in three steps. First, we find the frequency distortion in the degraded image. Second, we compute the deviation of this frequency distortion from an allpass response of unity gain (no distortion). Finally, we weight the deviation by a model of the frequency response of the human visual system and integrate over the visible frequencies. We demonstrate how to decouple distortion and additive noise degradation in a practical image restoration system.

A Regression-Based Family of Measures for Full-Reference Image Quality Assessment

“…In experiments, the following 16 IQA measures with publicly available objective scores, or source-code, for used benchmarks took part in the optimisation: VSI [15], FSIM [13], FSIMc [13], GSM [19], IFC [12], IW-SSIM [10], MAD [23], MSSIM [9], NQM [6], PSNR [38], RFSIM [16], SR-SIM [14], SSIM [8], VIF [11], IFS [41], and SFF [42]. It is worth noting that MAD is a multimeasure, but it was used in the optimisation due to its popularity and availability of the source-code.…”

“…One of the simplest techniques in this category is peak signalto-noise ratio (PSNR); noise quality measure (NQM) [6], in turn, uses a linear frequency distortion and an additive noise injection. Other measures which have been introduced in the last decade use: luminance and contrast distortions [7], structural information [8], [9], [10], statistical properties [10], [11], [12], phase congruency and image gradient magnitude [13], visual saliency maps [14], [15], Riesz-transform features [16], speeded-up robust features [17], local binary patterns [18], structure and contrast changes [19], inter-patch and intra-patch similarities [20], or fuzzy gradient similarity deviation [21].…”

Non-Shift Edge Based Ratio (NSER): An Image Quality Assessment Metric Based on Early Vision Features

“…Denote by the operator that counts the number of edge points in an NSE map. A ratio to measure the similarity between and , denoted by , is defined as (3) Clearly the range of pi is within [0, 1]. The more similar the two edge maps are to each other, the higher the value of pi is.…”

“…[1]. Later, some typical metrics, such as the Sarnoff JNDmetrix visual discrimination model (VDM) [2], the noise quality measure (NQM) [3], and the wavelet based visual signal to noise ratio (VSNR) [4] have been proposed. Instead of using the known HVS models, some IQA methods attempt to directly model the property of perceptual distortion of HVS, including the universal image quality index (UQI) [5], the Structural-SIMilarity (SSIM) index [6] and its multiscale counterpart MS-SSIM [7], the Information Fidelity Criteria (IFC) based on Natural Scene Statistics (NSS) [8], and the Visual Information Fidelity (VIF) [9].…”

No-Reference Image Quality Assessment through SIFT Intensity

“…There are a variety of HVS models, each of which can model parts of human vision (e.g. spatial resolution, temporal motion, color fidelity, color resolution, contrast and orientation sensitivity, frequency selectivity) [1,2,3,4,5,6]. By integrating with some HVS characteristics, the performance of some present metrics can be improved.…”