Variable bit-rate coding based on human visual system

Automatic image quality assessment has many diverse applications. Existing quality measures are not accurate representatives of the human perception. We present a hybrid image quality (HIQ) measure, which is a combination of four existing measures using an ‘n’ degree polynomial to accurately model the human image perception. First we undertook time consuming human experiments to subjectively evaluate a given set of training images, and resultantly formed a Human Perception Curve (HPC). Next we define a HIQ measure that closely follows the HPC using curve fitting techniques. The HIQ measure is then validated on a separate set of images by similar human subjective experiments and is compared to the HPC.The coefficients and degree of the polynomial are estimated using regression on training data obtained from human subjects. Validation of the resultant HIQ was performed on a separate validation data. Our results show that HIQ gives an RMS error of 5.1 compared to the best RMS error of 5.8 by a second degree polynomial of an individual measure HVS (Human Visual System) absolute norm (H 1) amongst the four considered metrics. Our data contains subjective quality assessment (by 100 individuals) of 174 images with various degrees of fast fading distortion. Each image was evaluated by 50 different human subjects using double stimulus quality scale, resulting in an overall 8,700 judgements.KeywordsImage QualityMean Square ErrorHuman Visual SystemImage Quality AssessmentOpinion ScoreThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

The bandwidth flexibility offered by the asynchronous transfer mode (ATM) technique makes it possible to select picture quality and bandwidth over a wide range in a simple and straightforward manner. A prototype model of a video codec was developed that demonstrates the feasibility of both variable bit rate (VBR) coding and user-selectable picture quality. The VBR coding algorithm is discussed and it is shown how a stabilized quality is achieved and how this quality and associated bandwidth can be selected by the user. How error propagation is limited to reduce the visibility of cell losses is also discussed. Interfaces with the ATM network are analyzed, with emphasis on decoder synchronization and absorption of cell delay jitter. The VBR codec offers very good picture quality for videophony applications at an equivalent load of 5.9 Mb/s. Picture quality remains relatively constant, even for heavy motion. >

The necessity of multi-spectral improved vision systems usage is shown. Image contrast enhancing algorithms are considered. The modification proposed allows to reduce the influence of artifacts in image areas with sharp brightness variety. Full-reference quality assessment of contrast enhancement algorithms application results is realized. Multispectral image dataset is created. The values of objective and subjective multispectral image fusion quality assessments and one subjective assessment are obtained. Correlation coefficients between subjective and objective assessments are calculated. It is shown that the Pearson product-moment correlation coefficient between objective and specified subjective image fusion quality assessments are not less than 0.54.

This paper describes a perceptual measure for still image compression system. Considering the fact that the conventional PSNR cannot sufficiently reflect the result of subjective assessment, other quality measure have been considered to design the variable bit-rate coders. Indeed, there is a growing interest for perceptual quality measure. Some works have been carried out in the field of still picture quality evaluation while trying to introduce some properties of the human visual system. In the recent literature there are roughly three properties that are identified as being useful. The best known, and generally most widely used properly, is the modulation transfer function of the human visual system. The other tow properties can be described as luminance masking and texture masking. A large number of image quality measures of this kind have been developed, with different degrees of success. In previous works, we provided a rigorous evaluation of metrics which take into account artifacts generated by compression method like JPEG. The results show that these metrics are highly correlated with the subjective quality grading but also depend on the complexity of the images under study. Then, we propose a new perceptual metric for still image compression based on multiresolution decomposition. It allows characterize image texture, better takes into account masking effect and don't depend on compression method.

An essential determinant of the value of digital images is their quality. Over the past years, there have been many attempts to develop models or metrics for image quality that incorporate elements of human visual sensitivity. However, there is no current standard and objective definition of spectral image quality. This paper proposes a reliable automatic method for objective image quality measurement by wavelet transform and Human visual system. This way the proposed measure differentiates between the random and signal-dependant distortion, which have different effects on human observer. Performance of the proposed quality measure is illustrated by examples involving images with different types of degradation. The technique provides a means to relate the quality of an image to the interpretation and quantification throughout the frequency range, in which the noise level is estimated for quality evaluation. The experimental results of using this method for image quality measurement exhibit good correlation to subjective visual quality assessments.

Measuring and reporting cardiac surgery quality: A continuing evolution

We have used the psychophysical methods described in the first paper of this series (Holliday and Ruddock, 1983) to determine selected spatial and temporal response characteristics of the ST1 and ST2 filters for subjects suffering visual defects. Data are given for 19 amblyopes, an albino and a hemianope, and comparison data are also given for a number of subjects with normal vision. The ST1 spatial responses for both the "normal" and "amblyopic" eyes of 12 convergent strabismic amblyopes are displaced to low spatial frequencies compared to the normal curve, which implies that there is a loss of fine spatial tuning. In all but one subject, the curve for the "amblyopic" eye peaks at a spatial frequency lower than that for the "normal" eye, thus the former deviates further from the normal pattern than the latter. The ST1 spatial responses of 6 refractive amblyopes are also displaced to the low frequency side of the normal curve, although on average the shift is smaller than in the case of the strabismic amblyopes. For each subject, the response curve of the "amblyopic" eye peaks at a lower spatial frequency than does that for the "normal" eye. ST1 spatial responses were measured for targets located up to 30 degrees off-axis along the horizontal meridian and sample data are given for one strabismic and one refractive amblyope and for two normal subjects. It is concluded from these data that the changes in the spatial responses associated with amblyopia do not simply reflect eccentric fixation of the target. The ST2 spatial response was measured for the "normal" and "amblyopic" eyes of 9 amblyopes (7 strabismic and 2 refractive). There is no significant difference between the average amblyopic response and that of normal subjects, and only in one case does the response for an "amblyopic" eye peak at a frequency lower than the peak frequency for normal vision. The ST2 temporal response for 9 amblyopes shows no systematic deviations from the normal response. For the albino, both the ST1 and ST2 spatial responses peak at around 0.3 cycles deg-1, and both curves are displaced considerably to the low spatial frequency side of the normal ST2 spatial response. The albino's ST2 temporal response is essentially normal. Measurements for the hemianope's "blind" hemifield under conditions appropriate to the isolation of the ST1 and ST2 spatial responses reveal no tuning curves. The ST2 temporal response for the "blind" hemifield, however, is of large amplitude, with a peak at 2 Hz, well below the normal frequency response peak. It is argued that the loss of fine spatial tuning which occurs in the ST1, but not the ST2, spatial responses of the amblyopes is consistent with the sequential organisation of these two filter classes proposed by Holliday and Ruddock (1983). Further, for the only two subjects whose ST2 spatial response curves are displaced to abnormally low frequencies (the albino and a strabismic amblyope) the ST1 spatial response is shifted to low spatial frequencies compared to the normal ST2 curve...

Image quality assessment has diverse applications. A number of image quality measures (IQMs) are proposed, but none is proved to be true representative of human perception of image quality. We have subjectively investigated spectral distance-based and human visual system (HVS)-based IQMs for their effectiveness in representing the human perception for images corrupted with white noise and Gaussian blur. Two sets of 160 images with various degrees of white noise and Gaussian blur are subjectively evaluated by 50 human subjects, resulting in 16 000 human judgments. On the basis of evaluations, image-independent human perception values are calculated. The perception values are plotted against spectral distance-based and HVS-based IQMs. The performance of quality measures is determined by graphical observations and polynomial curve fitting, resulting in best performance by HVS absolute norm and block spectral phase-magnitude error for white noise and Gaussian blur distortions, respectively. It is also observed that the performances of various quality measures differ for different noise distortions, suggesting the use of different quality measures for different noise types rather than a single universal IQM.

Ideally, a quality assessment system would perceive and measure image or video impairments just like a human being. But in reality, objective quality metrics do not necessarily correlate well with perceived quality [1]. Plus, some measures assume that there exists a reference in the form of an “original” to compare to, which prevents their usage in digital restoration field, where often there is no reference to compare to. That is why subjective evaluation is the most used and most efficient approach up to now. But subjective assessment is expensive, time consuming and does not respond, hence, to the economic requirements [2,3]. Thus, reliable automatic methods for visual quality assessment are needed in the field of digital film restoration. The ACE method, for Automatic Color Equalization [4,6], is an algorithm for digital images unsupervised enhancement. It is based on a new computational approach that tries to model the perceptual response of our vision system merging the Gray World and White Patch equalization mechanisms in a global and local way. Like our vision system ACE is able to adapt to widely varying lighting conditions, and to extract visual information from the environment efficaciously. Moreover ACE can be run in an unsupervised manner. Hence it is very useful as a digital film restoration tool since no a priori information is available. In this paper we deepen the investigation of using the ACE algorithm as a basis for a reference free image quality evaluation. This new metric called DAF for Differential ACE Filtering [7] is an objective quality measure that can be used in several image restoration and image quality assessment systems. In this paper, we compare on different image databases, the results obtained with DAF and with some subjective image quality assessments (Mean Opinion Score MOS as measure of perceived image quality). We study also the correlation between objective measure and MOS. In our experiments, we have used for the first image test set “Single Stimulus Continuous Quality Scale” (SSCQS) method and in the second “Double Stimulus Continuous Quality Scale” (DSCQS) method. The users, which are non-experts, were asked to identify their preferred image (between original and ACE filtered images) according to contrast, naturalness, colorfulness, quality, chromatic diversity and overall subjective preference. Test and results are presented.

Digitized hand-wrist radiographs: Comparison of subjective and software-derived image quality at various compression ratios

Measurement of the Spatial Frequency Response of Image Devices

A note on the use of subjective methods for assessing pig meat quality on the slaughterline

With the rapid advancement of haptic feedback technology for stylus pens, it has become increasingly important to accurately measure and evaluate the fidelity of haptic feedback. In this article, we propose a deep metric network framework based on triplet convolutional neural network (Triplet-CNN) to compare the haptic feedback fidelity of styluses. Unlike commonly used objective assessment methods, which provide a score for a single test sample, our approach focuses on comparing the haptic feedback fidelity between different styluses. This is because individuals tend to instinctively compare different options rather than assigning an absolute score to a single test sample. Our method can predict the proportion of users who consider one of the evaluated writing instruments to be more similar to the reference writing instrument than the other based on the vibrotactile signals generated during the pen-surface interaction. To train our network model, we collected acceleration data of pen-surface interactions at different writing speeds and pressures as inputs to the model. We then conducted pairwise comparative subjective experiments to construct a subjective assessment dataset as model outputs. To evaluate the performance of our model, we conducted cross-validation experiments. We used the correlation coefficient between subjective and objective assessment, mean absolute error (MAE), and prediction accuracy within a tolerance error as metrics to measure the consistency between the model’s predictions and the subjective assessment results. The experimental results demonstrate that our model’s predictions align more closely with subjective assessment than traditional signal similarity-based assessment metrics.

This dissertation proposes a novel gaze-driven approach for dynamic discomfort glare assessments as a first step towards understanding human responsive comfort with respect to daylight. The objective was to observe the natural gaze behaviour in relation to glare for office spaces with the conditions implicitly constrained by real world luminous conditions. In the existing visual comfort models human behaviour is not sufficiently considered. These models employ only subjective assessments, which lack an objective understanding of the factors affecting the perceptual mechanism of light-induced visual discomfort. They so far have not integrated the inter-dependencies of visual discomfort perception and human gaze responses and have been limited to a fixed-gaze assumption directed towards the office task area. In this dissertation, a gaze-driven approach is developed and adopted in the discomfort glare assessments. The assessments were done in a series of experiments in simulated office setting under different lighting conditions where participants’ gaze responses with means of mobile eye-tracking as well as their subjective assessments were recorded while monitoring photometric quantities relevant to visual comfort using high dynamic range luminance imaging. Integration of the luminance images coupled with eye tracking enabled us to obtain the gaze-centred luminance fields, which gives a better estimate of actual luminance values perceived by the eye, used as a basis to investigate the gaze direction dependencies of visual comfort. This PhD dissertation describes different stages of conception of this novel dynamic discomfort glare assessment method. In the experimental phase, two pilot studies were made for proper integration of the adopted methods and techniques into discomfort glare assessments. Development of several routines, algorithms and tools to identify and translate the gaze directions in order to derive the actual luminance field perceived by the participants were needed to achieve this goal. A final comprehensive experimental phase was realised to investigate gaze behaviour in response to light. As a first validation step, the gaze-driven approach was compared to the fixed-gaze approach. Then the effects of different luminance levels as well as different view outside the window on the dynamic shifts of the gaze were investigated.The developed approach demonstrates the need to integrate gaze direction patterns into visual comfort assessments, which move us beyond the existing assumption of a fixed-gaze direction towards a gaze responsive comfort.

In this paper, we propose an embedded variable bit-rate (VBR) audio coder to provide the fittest quality of service (QoS) and better connectivity of service for the ubiquitous speech communications. It has scalable bandwidth for narrowband to wideband speech signal, and embedded 8 32 kbit/s VBR corresponding to the network condition and terminal capacity. For the design of the embedded VBR coder, the narrowband signals are compressed by an existing standard speech coding method for the compatibility with G.729 coder, and then the other signals are compressed hierarchically on the basis of CELP enhancement and transform coding with temporal noise shaping (TNS) method. By the objective and subjective quality tests, it is shown that the proposed embedded VBR audio coder provides a reasonable quality compared with existing audio coders such as G.722 and G.722.2 in terms of mean opinion score (MOS) and perceptual evaluation of speech quality of wideband (PESQ-WB).