Quality Matrix in Terms of Software Quality Management - Literature review Example

Quality Matrix in terms of software quality management Name: Number: Course: Lecturer: Date: Abstract For the last couple of years, there is been some interest in the development of video coding technologies which has catapulted the creation and development of video communications. The need have a well defined mechanism for evaluating the quality of video is of paramount importance. This is because it is instrumental in the design of overall design of video communications. It is common knowledge that metrics which are simple and are used to measure energy like Peak Signal Noise Ratio (PSNR) are not efficient and reliable in measuring the degradation that are seen by the viewer. Recently, there has been a proposal of new video quality metrics that imitate the perception that humans want in a video since they produce results which are desirable and which are common to those that are produced by subjective methods. The new models that have been proposed have better accuracy than that found in PSNR method, their results are known to be consistent when the range of the data is used from the tests that are subjective have been carried out at a cross-section of videos. This paper will look at the features of these new models and how they are used to the well-known Hypothetical Reference Circuits (HRC) like video compression standards, bit-error transmissions and the packet losses. Table of Contents Abstract 2 Table of Contents 3 List of tables 4 Introduction 5 Objective of video quality metrics 5 Full-Reference quality metrics 6 MPQM tool 7 Reduced-Reference quality matrix 7 Video quality metrics 8 Video quality matrix experiment 10 Scenario I: significances of Codecs and format sizes 11 Scenario two: effects of the Bit-errors 13 Scenario three: effects of the video packet losses 13 Conclusion 16 References 17 List of tables Figure 1: Overview of QVM 8 Figure 2: Network topology of scenario 3 14 Figure 3: Video Metrics Results - scenario 3 15 Introduction The laying out of mechanisms that are used to assess the quality of video play a crucial role in the overall l design of video communication systems. As Tong, Heeger and Lambrecht (1999) argue, most work that has been done by the research community have dwelt mostly on the challenges and the problems of image quality assessment and it is only recently that the evaluation of quality of video been given some attention. There are several methods of assessing the quality of video. Among the many methods, the most reliable method of evaluating video quality is subjective evaluation. One reason for this is that human beings are the final receivers and they are the receivers in most situations. The first measurement is that of mean opinion score (MOS) that is used subjectively and is got from human observers; this measurement has been regarded as the most reliable form of measuring quality. Although this is the case, this method of measurement is so; it is known to be too inconvenient, slow and expensive when used by in most applications. The advantage of objective quality management is that they provide the designers of video and standard organisations a way of making meaningful and important quality evaluation without the need to convene convenor panel. The main objective of undertaking a research in video quality assessment is to come up with design quality metrics that are able to predict the quality of video automatically. This paper will first start by doing a research study on the current trends that define video metrics. The paper will then go ahead and analyse the capabilities of two representative examples when they are applied to different Hypothetical Reference Circuits (HRC) like bit-error transmissions and packet losses, and standards for compression of video. Each of the metric will be used to analyse the correlation that exists between predictions and the subjective ratings. Objective of video quality metrics Most of the video quality metrics standards that have been proposed in the literature make one assumption that there is availability of reference signal that is undistorted. Such video metrics are referred to as Full-Reference (FR) video quality metrics. It is important to note that in numerous practical video applications, the video that is used as the reference cannot be accessed. These are commonly referred to as No-Reference (NR) video quality metrics. There is another metric that is used to assess video quality. In this metric, the original signal of the video is not available completely. In this case, there are some extra features which are got from the original video and then transmitted to the system that is used for assessment as part of side information so that it aids to assess the quality of the video that has been distorted. This is what is referred to as Reduced-Reference (RR) video quality metrics. The whole issue of video assessment is still not well developed and is therefore still far from being a mature research topic. The group which is tasked with developing, standardising new methods of measurement used for quality of the video is the Video Quality Experts Group (VQEG). The development is done in phase. The first phase achieved minimal success. This phase entailed FR television video quality. Although this is the case, VQEG stills goes on working on the quality processes. Phase 2 of the process entails quality assessment for television, RR and NR that are the assessment criteria for television and multimedia. Full-Reference quality metrics In this category, the most used measurement standards include Mean Square Error (MSE), Signal to Noise Ratio (SNR) and Peak Signal to Noise Ratio (PSNR). These are commonly used by the scientific community because they are simple to calculate, have well-defined meanings, and are easy to deal with mathematically for the purposes of optimizations. Wolf et al. (1997) state that although this is the case, it is known that the metrics are not well suited to be used in the description of subjective degradation that is perceived by the viewer. This is the reason why there is a need to have definition of metrics which are able to take two factors. The first factor is the quality as perceived by the human system of viewing and the temporal relation that is present in the sequences of the video. Up to here, therefore, the video assessment quality metrics that belong to this family have been developed considering two physiological characteristics of the Human Visual System (HVS). In summary, then, an elaborative metric for assessing video quality will be able to implement the following HVS features: eye optics that are modelled by PSF of low-pass, retinal sampling that is not uniform, colour processing, adaptation of light (that is the masking of luminance), spatial frequency, contrast sensitivity, temporal frequency and analysis of orientation selective signal, contrast response saturation and pooling, and facilitation and masking. MPQM tool Developed by Van de Branden Lambrecht et al (1999), this tool is a representative of the HVS models. This is a proposal of Moving Picture Quality Metric (MPQM) and is consists of channel decomposition into four scales, four orientations and two streams which are temporal. The outputs of the resulting channel are then subtracted to create the error signal. Masking is then implemented by performing normalisation of the channel errors using the stimulus dependent visibility thresholds (which is equivalent to those that are found in still image quality assessment quality metrics). There is also a proposal for Motion rendering in quality assessment that has been proposed through the extension of MPQM; this is achieved by extracting motion information. As noted by Ghanbari and Pearson (1997), one factor that greatly affects HVS feasibility based video quality metrics is the complexity of computation. Although complex assessment methods may be used to model HVS accurately, the complexity in computation can be large in a prohibitive way in many platforms; this is especially true quality assessment methods for videos which has high resolutions. Due to the limitations that have been encountered in these methods and the lack of better assessment methods for these, other authors have proposed another alternative way to think about the assessment of quality of video using FV. The main function of the visual systems for human beings is to extract structural information from the field of viewing; the human viewing systems are well adapted for this. For this purpose, the measurement for structural distortions should serve as a good approximation of perceived image distortion. Reduced-Reference quality matrix These are metrics which are based on linear and non-linear transformations of distortion metrics with considerations of certain characteristics of image. In these metrics, some characteristics that are considered include block effects, fuzzy edges and distortions which are temporal. After this, a set of coefficients of linear or non-linear combinations is then created and defined. The main aim of this selection is to maximise the correlation that exists between the image characteristics and the results and outcome of a set of tests which are subjective. Video quality metrics Figure 1: Overview of QVM According to Wolf et al. (1999), the rest of the paper is based on linear and non-linear transformations of distortions metrics; this metrics is one example of video quality metrics and is based on localised spatial (SI) and the temporal (TI) activity features. There are three metrics that are used for comparison purposes; they are derived from the SI and TI features of the distorted and also of the reference videos. The reference video features are transmitted over the RR channel. Ghanbari and Pearson (1997) state that there was also an extension of the work where the quality of the video was developed so that it used reduced bandwidth features that are derived from spatial-temporal (S-T) regions of input which are processed and output video scenes. These characteristics show spatial detail, motion, and colour present in the sequence of the video. The activity of image ages, or spatial ingredients are characterised by spatial features. It is known that digital video systems can add edge, like edge noise and blocking, or reduce edges, like blurring. Temporal features also show characteristics the activities of temporal differences; they can also characterise temporal ingredients between frames which are successive of each other. Digital video systems can add motion, like error blocks, or reduce motion, like frame repeats. Feature of chrominance are used to characterise information of colour. Digital video systems are known to add colour information, like cross colour, or reduce information on colour, like colour sub-sampling. The parameters of either gain or loss is computed by way of doing a comparison of two parallel streams of feature samples; the samples are that one is from the input while another is from the output. The parameters of gain and loss are examined separately because each pair stream is used to measure completely different aspects of quality perception. The function that is used to calculate gain and loss tries to follow the perceptibility and the impairments by way of modelling perceptibility thresholds, error pooling and visual masking. A linear integration of the parameters is used to make an estimate of the subjective quality rating. The impairment metric that is dominated in the Video Quality Metric is defined by making use of these features. VQM has five structures with each structure being objective video quality model. These models are 1) General 2) Television 3) Video conferencing 4) Developer 5) PSNR. The general model was developed so that it will be used to be a general purpose VQM that will be used for video systems that cover a wide range of quality and bit rates. The television model was created optimised for television impairments (for example MPEG-2) while the third model, the video conferencing model was developed and optimised to be used for impairments of video conferencing, for example the H.263 and MPEG-4. The last model, the developer model, was developed for use with the same wide range of video quality and bit rates just like the wide range of video quality and bit rates as the general model but with added constraint of fast computation. The PSNR model was got from the traditional PSNR which is derived by using the following formula: VQMp = 1/(1+e0.1701*(PSNR-25.6675), 10 Read More