Other Free Encyclopedias » Online Encyclopedia » Encyclopedia - Featured Articles » Contributed Topics from P-T

Segmentation and Coding

based image video regions

Definition: Various coding schemes employ segmentation for different purposes, such as for extracting image regions in order to subsequently approximate the contours and image characteristics of those regions, or for object-based coding .

Segmentation is an important tool for image and video coding applications. Various coding schemes employ segmentation for different purposes, such as for extracting image regions in order to subsequently approximate the contours and image characteristics of those regions, or for object-based coding. The latter presents a framework for achieving improved coding efficiency without compromising the quality of at least the important parts of the image or video, such as a person’s face in a videoconference image sequence, particularly in very low bit-rate communications. One way of achieving this goal is by coding with lower quality the least significant parts of the sequence. Taking advantage of coding schemes like the aforementioned ones requires the use of a segmentation algorithm for partitioning the visual medium to spatial or spatio-temporal objects.

With respect to standards, the MPEG-4 International Standard supports the representation of video as a collection of objects, enabling object-based video coding; however, no International Standard (including MPEG-4) defines a set of segmentation tools that can be used for this or any other application. Thus, numerous segmentation approaches for efficient image and video coding have been proposed, considering various coding techniques.

In an object-based video compression scheme based on the derivation and efficient coding of motion boundaries is developed. It is based on initially identifying a small number of global movement classes. Then, regions of a spatial segmentation are assigned to each movement class and spatial regions are merged using various similarity metrics. Eventually, the boundaries of different motion classes are coded using an efficient asymmetric binary tree coding scheme.

In , two approaches to object-based content representation for coding are developed: a foreground/background segmentation method for stereo sequences and a face segmentation method for monoscopic video. Using these segmentation approaches, a constant-quality variable bit-rate (CQ-VBR) control algorithm is developed. It is shown that in low bit-rate applications, a significant reduction in the overall bit-rate can be achieved while maintaining the same visual quality of the region of interest (the foreground and face object, respectively), as compared to conventional frame-based coding.

A different approach is followed in, where the problem of efficient coding of still images overcoming the need for defining regions of interest is considered. A coding scheme based on quad-tree segmentation (i.e. a split and merge approach) is proposed, that can jointly encode similar neighboring segments. In doing so, this scheme achieves the optimal rate-distortion behavior for piecewise polynomial signals and is shown to be superior to conventional schemes not making use of a segmentation method.

Segmentation Evaluation [next] [back] Seebeck, Thomas Johann

User Comments

Your email address will be altered so spam harvesting bots can't read it easily.
Hide my email completely instead?

Cancel or