Statistical Evaluation of Visual Quality Metrics for Image Denoising

Noise is one of the main factors that degrade image visual quality. Assessment of perceptual quality of noisy images is of critical importance for imaging systems and image processing application. In this paper, we propose a framework that employs multi-layer neural networks to predict visual quality for noisy images. In particular, the proposed method utilizes quite simple features extracted in discrete cosine transform (DCT) domain for predicting values of full-reference visual quality metrics for noisy images. The neural networks are trained and tested on noisy images corrupted by additive white Gaussian noise (AWGN). The experimental results show that prediction is quite accurate and fast. Source codes of the proposed method and datasets will be available at https://github.com/ViA-RiVaL/Blind-Noisy-Image-Visual-Quality-Prediction.

To examine the performance of state-of-the-art wavefront-guided scleral contact lenses (wfgSCLs) on a sample of keratoconic eyes, with emphasis on performance quantified with visual quality metrics, and to provide a detailed discussion of the process used to design, manufacture, and evaluate wfgSCLs. Fourteen eyes of seven subjects with keratoconus were enrolled and a wfgSCL was designed for each eye. High-contrast visual acuity and visual quality metrics were used to assess the on-eye performance of the lenses. The wfgSCL provided statistically lower levels of both lower-order root mean square (RMS) (p < 0.001) and higher-order RMS (HORMS) (p < 0.02) than an intermediate spherical equivalent scleral contact lens. The wfgSCL provided lower levels of lower-order RMS than a normal group of well-corrected observers (p << 0.001). However, the wfgSCL does not provide less HORMS than the normal group (p = 0.41). Of the 14 eyes studied, 10 successfully reached the exit criteria, achieving residual HORMS wavefront error less than or within 1 SD of the levels experienced by normal, age-matched subjects. In addition, measures of visual image quality (logVSX, logNS, and logLIB) for the 10 eyes were well distributed within the range of values seen in normal eyes. However, visual performance as measured by high-contrast acuity did not reach normal, age-matched levels, which is in agreement with prior results associated with the acute application of wavefront correction to keratoconic eyes. Wavefront-guided scleral contact lenses are capable of optically compensating for the deleterious effects of higher-order aberration concomitant with the disease and can provide visual image quality equivalent to that seen in normal eyes. Longer-duration studies are needed to assess whether the visual system of the highly aberrated eye wearing a wfgSCL is capable of producing visual performance levels typical of the normal population.

Remote sensing images are subject to complex noise and distortions, which reduce their quality and can lead to loss of information. Visual quality metrics can be applied to automate and optimize remote sensing image processing. The problem of taking into account distortions typical for remote sensing data in existing test image databases is considered in this paper. The efficiency of 50 full-reference quality metrics for typical remote sensing distortions is estimated. The paper proposes a robust combined metric based on alpha-trimmed mean. The influence of linearization and the use of various correlation coefficients on the formation of a robust metric are studied. The effectiveness of the solution is tested using the cross-database approach based on the test image databases including TID2013, KADID10k, MDID and LIVE Multiply Distorted Image Quality Database.

Designing daylight-based spaces has gained increasing attention due to its numerous benefits and alignment with global sustainability standards. However, limited research has focused on how architectural layouts affect daylight distribution and visual quality, particularly in educational environments. This research aims to address this gap by integrating spatial layout analysis with daylighting and visual quality metrics to enhance daylight benefits across seven school plan layouts as case studies. Using a systematic review and simulation-based methodology, daylight performance is evaluated with Useful Daylight Illuminance (UDI up, UDI, UDI low) and Daylight Autonomy (DA) using the Honeybee plugin and Radiance engine in Rhinoceros software. Visual quality is also explored through Isovist and Depth Map analysis, focusing on metrics including connectivity (C), Isovist area (A), drift angle (DA), Maximum radial line (RL(L)), Minimum radial line (RL(S)), and Visual Mean Depth. Simulation validation is conducted through a lux meter in a real classroom in Tehran, Iran, along with SPSS analysis to explore the accuracy and correlations between the results. Findings reveal that well-designed interior layouts significantly enhance both daylight distribution and visual quality. School with layout C, by higher Drift Angle (DA) and longer visual range (Max radial) can provide a dynamic and open environment, outperforming other layouts in both daylighting (DA = 52.93%, UDI = 68.17%) and visual quality metrics (Drift Angle ≈ 220, Connectivity ≈ 600). Layouts F and E also perform well, while Models D and G show less daylight performance (UDI-low ≈ 31–34%). The outcomes emphasize the importance of an integrated design approach including DA, UDI, Isovist metrics, and plan layouts, offering a new framework to enhance daylight benefits in educational spaces. It also provides valuable insights to improve architectural schools’ designs, contributing to more sustainable, visually enriched, and energy-efficient learning environments.

Textures are being extensively used in next-generation video coding, image restoration and computer graphics wherein the stochastic and perceptual properties of textures are exploited. Given the prominence of texture content in image and video processing applications, texture quality assessment and objective quality metrics geared specifically towards texture are of particular importance. This paper outlines an extensive subjective study in which a total of 340 distorted images extracted from 10 reference images were evaluated by a group of 20 human subjects. The subjective Mean Opinion Scores (MOS) thus obtained from the study were used to evaluate several existing mainstream full reference quality metrics with the help of a recently developed quality assessment framework (IVQUEST). Based on the results obtained, it can be concluded that traditional objective metrics do not perform well on texture images and there is a need for specialized texture quality metrics.

To improve the prediction accuracy of visual quality metrics for video we propose two simple steps: temporal pooling in order to gain a set of parameters from one measured feature and a correction step using videos of known visual quality. We demonstrate this approach on the well known PSNR. Firstly, we achieve a more accurate quality prediction by replacing the mean luma PSNR by alternative PSNR-based parameters. Secondly, we exploit the almost linear relationship between the output of a quality metric and the subjectively perceived visual quality for individual video sequences. We do this by estimating the parameters of this linear relationship with the help of additionally generated videos of known visual quality. Moreover, we show that this is also true for very different coding technologies. Also we used cross validation to verify our results. Combining these two steps, we achieve for a set of four different high definition videos an increase of the Pearson correlation coefficient from 0.69 to 0.88 for PSNR, outperforming other, more sophisticated full-reference video quality metrics.

A holistic modeling for QoE estimation in live video streaming applications over LTE Advanced technologies with Full and Non Reference approaches

While Denoising is an extensively studied task in signal processing research, most denoising methods are designed and evaluated using readily processed image data, e.g. the well-known Kodak data set. The noise model is usually additive white Gaussian noise (AWGN). This kind of test data does not correspond to nowadays real-world image data taken with a digital camera. Using such unrealistic data to test, optimize and compare denoising algorithms may lead to incorrect parameter tuning or suboptimal choices in research on real-time camera denoising algorithms. In this paper we derive a precise analysis of the noise characteristics for the different steps in the color processing. Based on real camera noise measurements and simulation of the processing steps, we obtain a good approximation for the noise characteristics. We further show how this approximation can be used in standard wavelet denoising methods. We improve the wavelet hard thresholding and bivariate thresholding based on our noise analysis results. Both the visual quality and objective quality metrics show the advantage of the proposed method. As the method is implemented using look-up-tables that are calculated before the denoising step, our method can be implemented with very low computational complexity and can process HD video sequences real-time in an FPGA.

The paper describes a new image database HTID for verification and training of no-reference image visual quality metrics. The database contains 2880 color images of size 1536×1024 pixels cropped from the real-life photos produced by the mobile phone cameras with various shooting and post-processing settings. Mean opinion scores for images of the database are obtained. Peculiarities of the database are considered. A comparative analysis of the state-of-the-art no-reference image visual quality metrics is carried out. It is shown that the proposed database takes its own unique place in the existing image databases and can be effectively used for metrics’ verification.

In this paper, we discuss how to assess video Quality of Experience (QoE) with image saliency and texture information extracted from original and distorted video sequences. Based on this information, we proposed two categories of full-reference quality metrics. The first category of metrics considers spatial distortions measured by MSE and SSIM in terms of saliency weighted images, saliency maps and texture maps. The second category of metrics are constructed by considering temporal distortions, which also includes the above three aspects. Then the temporal MSE between original and distorted video sequences is computed frame-by-frame. Thus altogether 9 metrics are obtained. With these metrics and subjective MOS from both the LIVE dataset and our own dataset, we conduct Neural Net fitting to measure the performance. Finally the detailed comparisons with mainstream models verify the effectiveness of the proposed model.

Super-multiview content with high angular resolution: 3D quality assessment on horizontal-parallax lightfield display

Background: During clinical subjective refraction, sphere, cylinder and axis are adjusted to achieve the best corrected acuity using patient responses. Certain visual quality metrics are highly correlated with acuity, potentially allowing better objective refractions by objectively answering the question “which is better, 1 or 2.”

Color images formed by modern digital cameras are often noisy, especially if they are captured in bad illumination conditions. This makes desirable to remove the noise by image pre-filtering. A specific feature of the noise observed for the considered application is that it can be spatially correlated. Filters to be applied have to effectively suppress noise introducing only negligible distortions into processed images. Moreover, such filters have to be fast enough and tested for a variety of natural images and noise properties. Another specific requirement is that a visual quality of processed images has to be paid a specific attention. To carry out intensive testing of some denoising approaches, a recently designed database TID2008 of distorted images provides a good opportunity since it contains 25 different images corrupted by i.i.d. and spatially correlated noise with several levels of variances. Taking into account the known fact that the color components are highly correlated, both modern 2D (component-wise) and 3D (vector) filtering techniques are studied. It is demonstrated that the use of 3D filters that allow exploiting inter-channel correlation provides considerably better results in terms of conventional and visual quality metrics. It is also shown how 3D filter based on discrete cosine transform (DCT) can be adapted to a spatial correlation of noise. This adaptation produces sufficient increase of the filter's efficiency. Examples of filter's performance are presented.

Many visual quality metrics are used to measure the quality of lossy compressed images and videos, and are integrated in the rate-distortion optimization of hybrid video codecs. However, most of the metrics focus on the average objective quality in a picture. In certain applications, like medical image processing, the maximum absolute error should be more weighted. In this paper, the rate-distortion optimization of HEVC is extended by integrating this error metric. Thus, rate, average error, and maximum absolute error are jointly optimized. Furthermore, a weighting factor α is included into the calculation of the optimization for balancing the ratio between average and maximum absolute error. For HEVC intra with $\alpha =0.25$ an average maximum absolute error reduction of −25.63 can be achieved, while the bitrate increases slightly by 0.59%. Furthermore, the visual quality of the medical volumes improves and the data fidelity increases, i.e. less block artifacts appear and less structure disappear.

To compare ocular and internal aberrations after femtosecond laser anterior capsulotomy and continuous curvilinear capsulorrhexis in cataract surgery. In this prospective study, anterior capsulotomy was performed during cataract surgery with an intraocular femtosecond (FS) laser (Alcon LenSx Inc) in 48 eyes. As a control group, continuous curvilinear capsulorrhexis (CCC) was performed in 51 eyes. Wavefront aberrometry, corneal topography, and objective visual quality were measured using the OPD-Scan (NIDEK Co Ltd). Vertical and horizontal tilt, coma, and visual quality metrics were evaluated separately to determine whether the source of aberrations was ocular or internal. Main outcome measures included postoperative residual refraction, uncorrected and corrected visual acuities, ocular and internal aberrations, Strehl ratio, and modulation transfer function (MTF). No statistically significant differences were noted between the FS and CCC groups, respectively, in postoperative sphere (-0.60 ± 1.50 vs -0.50 ± 1.40 diopters [D]), postoperative cylinder (1.30 ± 1.01 vs 1.10 ± 1.10 D), uncorrected distance visual acuity (0.86 ± 0.15 vs 0.88 ± 0.08), or corrected distance visual acuity (0.97 ± 0.08 vs 0.97 ± 0.06). The FS group had significantly lower values of intraocular vertical tilt (-0.05 ± 0.36 vs 0.27 ± 0.57) and coma (-0.003 ± 0.11 vs 0.1 ± 0.15), and significantly higher Strehl ratios (0.02 ± 0.02 vs 0.01 ± 0.01) and MTF values at all measured cycles per degree, compared to the CCC group. Capsulotomy performed with an intraocular FS laser induced significantly less internal aberrations measured by the NIDEK OPD-Scan aberrometer compared to eyes that underwent CCC, which may result in better optical quality after the procedure.