Content-prioritised video coding for British Sign Language communication.

摘要: Video communication of British Sign Language (BSL) is important for remote interpersonal and the equal provision services deaf people. However, use video telephony conferencing applications BSL limited by inadequate quality. a highly structured, linguistically complete, natural language system that expresses vocabulary grammar visually spatially using complex combination facial expressions (such as eyebrow movements, eye blinks mouth/lip shapes), hand gestures, body movements finger-spelling change in space time. Accurate places specific demands on visual media which must compress image data efficient transmission. Current compression schemes apply methods to reduce statistical redundancy perceptual irrelevance based general model Human Visual System (HVS) sensitivities. This thesis presents novel coding developed achieve conflicting requirements high quality coding. Novel prioritising content optimised are exploit HVS spatial temporal response mechanisms users (determined Eye Movement Tracking) characteristics content. The implement an accurate foveation, applied domains, at pre-processing stage current standard-based (H.264). Comparison performance standard systems, evaluation this thesis, demonstrates improved perceived low bit rates. users, broadcasters service providers benefit from perception over range available transmission bandwidths. research community benefits new approach optimisation better understanding needs