Semantic Transcoding of Videos by Using Adaptive Quantization

Abstract

This paper proposes the use of an approach of video transcoding driven by the video content and provided with the adaptive quantization of MPEG standards. Computer vision techniques can extract semantics from videos according with users interests: the video semantics is exploited to adapt the video in order to meet the device’s capabilities and the user’s requirements and preserve the best quality possible. Well assessed video analysis techniques are used to segment the video into objects grouped in classes of relevance to which the user can assign a weight proportional to their relevance. This weight is used to decide the quantization values to be applied in the MPEG-2 encoding to each macroblock. A modified version of the PSNR (Peak Signal-to-Noise Ratio) is used as performance metric and comparative evaluation is reported with respect to other coding standards such as JPEG, JPEG 2000, (basic) MPEG-2, and MPEG-4. Experimental results are provided on different situations, one indoor and one outdoor.