CONTOUR APPROXIMATION FOR FASTER OBJECT BASED TRANSCODING WITH HIGHER PERCEPTUAL QUALITY
Oleg V. Komogortsev, Javed Khan · 2004
Object aware video rate transcoding can significantly improve the perceptual quality of relatively low bit-rate video. However, precise object detection in arbitrary video scenes is computationally challenging. This paper presents an interesting experimentation with a live eye-gaze tracker which suggests that object detection, particularly for perceptual encoding, may not have to be precise. Indeed, a simple approximation can not only reduce the complexity of the detection process, but also result in improved perceptual quality and yield very fast transcoding algorithms.