Ing. Adam Zizien

This paper introduces a public (OMNIQAD) database of omnidirectional (360 degrees) images distorted by various ways and techniques, which can serve as a basic source for various research in the field of objective and subjective quality assessment of an immersive multimedia content. Compression-based distortions were applied using traditional JPEG and several emerging image compression standards, for instance, HEIC, AVIF and JPEG XL. Other types of distortion included Gaussian and impulse noise. The database offers eleven reference omnidirectional images of different character (374 distorted images in total). For the evaluation of distorted and compression-based corrupted images, several objective quality metrics were adopted. Link to a repository containing the resulting objective scores and the database of images are included in this paper.

Recent developments in the field of plenoptic imaging have directly coincided with standardisation efforts. Such efforts are in no small part reliant on quality evaluations. For traditional imaging modalities, such as static images and videos, subjective evaluations have played a very important part. The goal of most objective quality evaluations has been the correlation performance with subjective results. However, plenoptic modalities enable applications that rely more on the objective rather than subjective quality. As such, it is necessary to examine the performance of currently used objective quality metrics. The paper provides a study on the predictive abilities of objective quality metrics on the quality of light field disparity map estimation. The study measures the influence of distortions on light field disparity estimations; furthermore, the study examines how well can objective quality metrics predict the quality of the resulting disparity maps. None of the tested metrics showed satisfactory prediction performance. Thus, two metric fusion algorithms were also examined to see if either can further improve prediction accuracy. The metric fusion algorithms showed that the prediction can be significantly improved. A novel objective quality metric that would correlate well with both subjective and objective results could thus be introduced.

Plenoptic modalities, in general, have seen a steady increase in research activity in recent years. Advances in this field pose additional demands on the development of new compression algorithms, which could be used for efficient delivery and storage of the acquired data. Subjective assessments are a key part of the development process. In this work, we examine the differences between two passive methodologies for the subjective assessment of light field images; a rating methodology, introduced in our previous work, and a ranking methodology. Additionally, we analyse the impact and viability of cross-content evaluations on the results of the assessment. The resulting data are made publicly available and can be used to design new compression algorithms, subjective assessment methodologies, or objective quality metrics.

An increase in research activity around plenoptic content can be seen in recent years. As the communities around the different modalities grow, so does the demand for publicly available annotated datasets of suitable content. The datasets can be used for a multitude of purposes, such as to design novel compression algorithms, subjective evaluation methodologies or objective quality metrics. In this work, a new publicly available annotated light field image dataset is presented. The dataset consists of scenes corrupted by state-of-the-art image and video compression algorithms (JPEG, JPEG 2000, BPG, VP9, AV1, AVC, HEVC), noise, and geometric distortion. For the subjective evaluation of the included scenes, a modified version of the Double Stimulus Impairment Scale (DSIS) methodology was adopted. The views of each scene were organized into a pseudo-sequence and played to the observers as animations. The resulting subjective scores, together with additional data, are included in the dataset. The data can be used to evaluate the performance of currently used visual quality metrics as well as for the design of new ones.