Low complexity binary description wavelet codec
Yiliang Bao, C.‐C. Jay Kuo · 1999
This paper presents a new coding algorithm, called the low complexity binary description (LCBiD) wavelet coder, which has an extremely low implementation complexity yet very rich features. LCBiD encodes wavelet coefficients in blocks of fixed size (16,16). Wavelet coefficients of a block are only taken from a single subband. No inter-subband correlation is exploited so that no specific requirement is imposed on how wavelet coefficients are generated. Each block is encoded using binary context-based bitplane coding. A (16,16) block is usually much smaller than a subband, and a block can become significant much later than a subband. A block skipping technique is employed to bypass one layer of zeros if a block is insignificant at the current quantization threshold. In other words, a separate bitstream is generated to describe the significance of blocks. Because much less zeros are encoded, both the coding speed and efficiency are increased greatly. A separate arithmetic coder set is used for each bitplane, and all coders are reset at the beginning of each subband. Thus, each bitplane of every subband in the compressed file can be directly accessed, and decoded almost independently. The only constraint is that all bitplanes of one subband must be decoded in the correct order. In order to reduce the overall number of arithmetic coders, very simple coding contexts are defined. The image codec has a very good coding performance, and the compressed file is quality and resolution scalable, and resilient to transmission errors.