Background mosaicking for low bit rate video coding

Abstract

This paper proposes a new technique to build a background memory based on mosaicking. More precisely, the technique rst identi es background and foreground regions based on local motion estimates. Camera motion is then estimated on the background by applying a parametric global motion estimation. Finally, after compensating for camera motion, the background content is temporally integrated in longterm memory. The method leads to high coding performances and allows for content-based functionalities.

Multimedia Systems, Standards, and Networks

“…Indeed, a video sequence can be represented by a sprite and warping parameters [23]. This concept is similar to the mosaicking techniques proposed in [27,28] and to the layered representation introduced in [29]. Therefore, sprite coding achieves a high coding efficiency.…”

Recent developments in panoramic image generation and sprite coding

“…Sprite Coding [31], [32], [33], [34] was developed and standardized in MPEG-4 some years ago. The main idea of this coding approach is to segment the video content of the input video into two parts, foreground object and background object, in a pre-processing step.…”

Lie Group Transformation Models for Predictive Video Coding

“…Even though the temporal correlation in 3D scenes arises from 3D rigid or non-rigid motions, which translate into affine, perspective or more complicated transforms on the image plane, video compression techniques remain limited to block-based motion estimation that exploits only correlation due to translational motion. More complicated transformation models such as those incorporating perspective or affine affine transformations stay limited to global predictive coding (e.g., background warping) [1], object and region-based coding [2], or motion refinement [3]. One of the main reasons for this is that the bit-rate overhead for encoding the transform parameters becomes prohibitive when the size of the blocks decreases.…”