Record:   Prev Next
作者 Turney, Robert D
書名 An adaptive quality metric for multimedia, video and imaging communications
國際標準書號 0542043149
book jacket
說明 96 p
附註 Source: Dissertation Abstracts International, Volume: 66-03, Section: B, page: 1641
Supervisor: Ali M. Reza
Thesis (Ph.D.)--The University of Wisconsin - Milwaukee, 2005
A new objective quality metric is proposed for multimedia, video and imaging communications. In recent times the deployment of systems that transport digital media has increased enormously. The devices now used in everyday life which include digital imagery are plentiful. From digital cameras to cellular telephones we find digital images and video being utilized and transported by wired and wireless systems. It is the focus of this work to provide an adaptive quality metric for the digital imagery so that assessing the overall quality of the human experience in digital imaging products can be made
The state of the art for objective quality metric is currently Peak Signal to Noise Ratio (PSNR). In PSNR the variance between the original image or video and a degraded image or video is calculated and the result given in dB. The reason for degraded imagery stems from the need to transport the imagery over wired and wireless channels. To have efficient use of the available bandwidth in digital communications, digital compression is utilized and artifacts arise due to the digital compression algorithm processing. Correlation between PSNR values and expert viewers give a measure of the relevant goodness of the PSNR metric. It is commonly accepted that subjective metrics through the use of expert viewers achieve better results than objective metrics. This work will compare with subjective data and strives to achieve better correlation than PSNR
In this work KRT_image and KRT_video metrics are proposed. The first step in the KRT_image algorithm involves performing a wavelet transform on the original and degraded imagery. In the wavelet domain regions of interest are identified through use of a local variance and a statistical F-test applied to the ratio of two local variances. Once the image areas are identified, different regions are weighted according to the relative importance to the human visual system and the weighted PSNR is calculated in the wavelet domain to achieve the KRT value in dB. The KRT_video algorithm is extended to take into account active versus non-active regions of interest. The results here show the adaptive KRT metric is superior to PSNR
School code: 0263
Host Item Dissertation Abstracts International 66-03B
主題 Engineering, Electronics and Electrical
Computer Science
Alt Author The University of Wisconsin - Milwaukee
Record:   Prev Next