S. Rao and N. Jayant (USA)
Telehealth, Region of Interest (ROI), Extended Region of Interest (EROI), Mathematical losslessness (M-lossless), Perceptual losslessness (P-lossless), Motion compensated Temporal Filtering (MCTF).
There is an increased need for high quality transmission of video over wireless networks for rapidly growing applications such as telehealth. We propose an algorithm for achieving true optimization between total bitrate and total quality. This is achieved through the notion of a diagnostically important Region-of-Interest (ROI) that is transmitted at high quality, and remaining regions (non ROI) that are transmitted at acceptable quality. We introduce a methodology for providing a less abrupt and more scalable distribution of quality as a function of space through the notion of an Extended-Region-of Interest (EROI). By efficient bitrate allocation between the ROI, EROI, and non-ROI, we show how the ROI and EROI can have high quality, and the non-ROI at least acceptable quality, even with bounds on available total bitrate. We apply two abstractions for high quality – Mathematical losslessness (M-lossless) and Perceptual losslessness (P-lossless). Within this framework, we extend our method to noisy video inputs and use Motion Compensated Temporal Filtering (MCTF) to optimally process such video signals.
Important Links:
Go Back
