|
|
|
|
Final Year Project
|
 |
MPEG-1 | ||
|---|---|---|---|
|
|
MPEG-2 | ||
|
|
MPEG-4 |
MPEG-1 is intended for coding of video of SIF 1 resolution for digital storage media up to about 1.5M bits/sec. This standard aims at the compressed representation of progressive video and its audio on various digital storage media, such as video compact disk (CDs) and optical drives.
MPEG-2 is intended for generic coding of moving pictures and associated audio. This standard allows compression of both progressive and interlaced video at various data bit-rated from about 1.5M bits/sec to more than 60M bits/sec, enabling application ranging from home entertainment quality video up to HDTV. The interesting applications of MPEG-2 technology are video-on-demand, digital TV/HDTV broadcasting and multimedia video.
For the MPEG-4 technology, the first version is announced at the end of 1998, and the second version is approved at the end of 1999. It intended for compression of full-motion video, and the bit rate of about 9-40k bits/sec. The applications are interactive multimedia and video telephony.
Moreover, MPEG-4, to allow access to contents of a picture, has introduced a new concept of Video Object planes (VOPs). The idea of VOP is that each video frame can be segmented into a number of arbitrary shaped image regions and encoded individually. By using this technology, it can achieve a higher compression ratio.
In this project, I am emphasis on two parts. They are image pre-processing and object based video compression system; these two applications are implemented using Visual C++.
For the image pre-processing part, all the edges within an image are extracted using Sobel Edge Enhancement Algorithm. In here, both vertical and horizontal mask are implemented in order to perform spatial filtering.
After extract all the edge information from an image, Smooth Algorithm is applied to blur the edge in order to attract snakes from a fairly large distance away.
For the object based video compression system, the shape-coding algorithm and texture-coding algorithm have been implemented.
In shape coding, based on the snake algorithm, chain code shape coding technique is used to find out the contour of an arbitrarily shaped object. In texture coding, the boundary of an object is encoded using extended-interpolation DCT, and the others are encoded using JPEG technology. This method uses DCT compression of 8x8 blocks of pixels, followed by quantization of the DCT coefficients and entropy encoding of the result.
After the compression system is completed, the following parameters should be measured and compared. They are compression ratio, signal-to-noise ratio, and bit allocation of background and object.
