The area of mobile communications has evolved in the last years and with the introduction of 3rd generation cellphones users can use features like video calls and mobile television. There are various bottlenecks in this kind of communication that introduce distortions to the transmitted video; therefore it is desirable to measure the video quality of these services. A subjective measurement takes long time to conduct, time which often isn’t available. Hence objective measurements with good correlation with the subjective quality evaluations are needed. The purpose of this master thesis is to develop a proof-of-concept application which objectively measures video quality.
Traditional video quality algorithms are based on error-sensitivity. The human visual system on the other hand is adapted to extract structural information from the viewing field. In this thesis the Structural Similarity Index Measurement (SSIM) algorithm is used. This algorithm is based on structural distortions and considers image degradations as perceived structural information loss. It is a full reference algorithm and thus needs access to the reference sequence in order to generate a quality rating. To be able to test the SSIM algorithm two video sequences were selected; one of a woman talking on the phone and the other with a rotating chessboard. Different distortions were then applied to the sequences to get a testing material. In the end ten distorted sequences from each reference were chosen, making a total of 20 sequences.
A subjective testing was conducted to get a reference quality measure for all generated video sequences. In total, there were 45 participants from Neava in Luleå and Ericsson in Lund in the testing which in the end rendered a mean opinion score (MOS) for each of the 20 sequences. Later the MOS from the subjective tests were compared to the objective quality rating from the SSIM algorithm.
In order to obtain the objective quality ratings an application prototype was developed. The prototype implements the SSIM algorithm and displays the quality for each frame individually in a graph and as a mean of the entire video sequence. In addition the application offers dual playback of both the reference and deformed sequences at the same time, and for demonstrational purposes there is also a graph which can show for instance block errors or received signal strength. These parameters can then be compared to the video quality rating. By then examining the results from the subjective- respectively objective tests the thesis hopes to answer questions about the reliability of objective measurements. The results look promising, but as discussed in the thesis further improvements can be made in different areas to improve the quality ratings.
Author: Moberg, Johan
Source: Lulea University of Technology
Download Link: Click Here To Download This Report (PDF)
Reference URL: Visit Now