UC Irvine

In this thesis, we work on quality estimation and delivery of visual content with different spatial resolutions. First, we study the quality estimation of images with different spatial resolutions. Estimating the quality of the visual content accurately is crucial in providing satisfactory multimedia communication. State of the art visual quality assessment approaches are effective when the input image and the reference image have the same resolution. However, finding the quality of an image that has spatial resolution different than that of the reference image is still a challenging problem. To solve this problem, we develop a quality estimator (QE) which computes the quality of the input image without resampling the reference or the input images. We begin by identifying the potential weaknesses of previous approaches used to estimate the quality of experience. Next, we design a QE, called Multiscale Image Quality Estimator (MIQE), to estimate the quality of a distorted image with a lower resolution compared to the reference image. We also propose a subjective test environment to explore the success of the proposed algorithm in comparison with other QEs. When the input and test images have different resolutions, the subjective tests demonstrate that in most cases the proposed method works better than other approaches. In addition, the proposed algorithm also performs well when the reference image and the test image have the same resolution.

Second, we examine the quality estimation of videos with different spatial resolutions. Full-reference video QEs either resize the distorted input video or the reference video to compute the quality when these videos have different spatial resolutions. This resizing operation causes several limitations. MIQE overcomes those limitations for images but it does not consider the temporal characteristics of video. We develop a full-reference video quality estimator that integrates MIQE with the motion information to estimate the quality of the distorted video without resampling the reference or the test videos. We also perform subjective tests to compare the proposed algorithm with the existing QEs. In these tests, the reference and the input videos are displayed at their native resolutions. Test results show that the proposed algorithm outperforms other QEs when the reference video and the input video have different spatial resolutions. We have also evaluated the performance of the approach using the Scalable Video Database.

Third, we work on the challenge of using a perceptual quality estimator to perform optimum multicasting of videos to the devices with different spatial resolutions. We specifically focus on the complexity of the optimum perceptual multicasting. The complexity increases due to usage of scalable video coding in combined scalability mode and perceptual quality estimators. Using combined scalability increases the number of scalability options, so we need to perform multi-criteria optimization. As a result of the simulations we have performed, we have observed that multi-criteria optimization is not necessary in the low bitrate region, and we propose an algorithm to reduce the complexity of the optimization notably for this region.