4. WAVELET BASED VIDEO COMPRESSION WITH ARITHMETIC CODING

           In this study a video compression algorithm has been developed. Each part of this algorithm has been implemented individually. Then, all these parts put together to compress a real video. Decompression algorithm is the inverse of this technique except quantization. In decompression de-quantization is not needed.

           Video compression algorithm contains four sections. One is the wavelet transform section. In this process a video frame is decomposed to its subbands. Second one is quantization. A new semi-uniform quantization algorithm has been developed. This process increases the occurrence of zeros and decreases the occurrence of non-zero coefficients. This new algorithm is more accurate than uniform quantizers and less complex than non-uniform quantization algorithms. Thirdly, ZTE coding algorithm is used to reduce the number of coefficients. According to ZTE coding rule many ZTE symbols are generated representing wavelet coefficients. Finally, arithmetic compression is applied to the symbols and all remaining non-zero coefficients.

4.1 Video Format

           Compressing a frame is the first step in compression scheme. All above sections describe how to compress a frame. One frame’s encoded structure is shown in Figure 15. Symbol length, coefficient length, ZTE symbols and non-zero coefficients are put together to form one block.

           However, in video compression there are many frames waiting to compress. To improve compression accuracy first three steps (DWT, Quantize, ZTE) are applied for each frame and arithmetic coding is applied for a group of frames. Deciding how many frames will be put together is determined by the user. This is done by getting a parameter KeyFrame. If this is set to 5, five frames will be sent to the arithmetic coder to compress. The structure of this process is in Figure 16. All five frames, in this example, are compressed once in arithmetic coder.

Figure 15. ZTE coded one frame structure.
 

 Figure 16. One big block for one KayFrame size.

Compressing all frame groups make a compressed video. The final compressed video file looks like as in Figure 17.
 

 Figure 17. Compressed video file.