130 likes | 360 Views
“P OLITEHNICA ” U NIVERSITY OF T IMIȘOARA FACULTY OF ELECTRONICS AND TELECOMMUNICATIONS DEPARTMENT OF COMMUNICATIONS. DIPLOMA THESIS. VIDEO QUALITY ASSESSMENT IN MOBILE NETWORKS. Coordinate, Assoc. Prof. Eng. Dr. Eugen Mârza. Graduate, Dragoș-Florin Iancu. Timișoara 2010. Contents.
E N D
“POLITEHNICA” UNIVERSITYOF TIMIȘOARAFACULTY OF ELECTRONICS AND TELECOMMUNICATIONSDEPARTMENT OF COMMUNICATIONS DIPLOMA THESIS VIDEO QUALITY ASSESSMENT IN MOBILE NETWORKS Coordinate, Assoc. Prof. Eng. Dr. Eugen Mârza Graduate, Dragoș-Florin Iancu Timișoara 2010
Contents • Introduction • Human Visual System (HVS) • Mobile Video Streaming Principles • Quality Metrics • Conclusions Dragoș Iancu
Introduction Topics of interest: • vision modeling in the framework of video quality assessment • analysis and evaluation of different video quality assessment methods and algorithms over a mobile network • determining the one that performs best while correlating well with subjective assessments • error detection and concealment (artifacts) Dragoș Iancu
Human Visual System • HVS: • very complex • highly adaptive • not equally sensitive to all stimuli • visual information processed on different pathways and channels • color perception based on different spectral sensitivities of photoreceptors => characteristics of the HVS used in the design of vision models and quality metrics Dragoș Iancu
Mobile Video Streaming Principles • Video service request in a mobile network • wireless=error prone environment • limited bandwidth means low resolution => loss of 1 packet is a considerable loss of information • real-time => retransmission impossible Dragoș Iancu
Subjective & Objective Quality Assessment Dragoș Iancu
Distortion Artifacts • Blocking artifacts • Ringing artifacts • Clipping • Noise • Contrast • Sharpness • Blurriness • Jerkiness • Mosquito Noise • Shimmering • Network errors • Post-processing errors Dragoș Iancu
Quality Metrics a) MOS (Mean Opinion Score) Subjective test Dragoș Iancu
Quality Metrics b) PSNR (Peak Signal to Noise Ratio) Dragoș Iancu
Quality Metrics c) SSIM (Structural Similarity) Dragoș Iancu
Quality Metrics • SSIM simple and multiscale plugins in ImageJ Dragoș Iancu
Conclusions 1. Best results with PSNR metric 2. SSIM best mimics the HVS, however has limitations 3. MOS delivers valuable information for: • optimization • benchmarking 4. methods dependent on: • the correct receival of motion vectors • the presence of scene cuts or fast movement 5. systematic video quality assessment approaches developed to increase flexibility Dragoș Iancu
Thank You! Dragoș Iancu