Subband Lossy and Lossless Coding

提出了两种有损和无损图像编码方法。一种是基于小波变换的，由全通滤波器并联组成的双通道IIR滤波器组来实现。这种结构非常简单，计算复杂度较低，但为了获得良好的性能，每种模式的参数都是相互独立的。另一种是基于重叠双正交变换，它由整数DCT和提升步长组成，以克服上述问题。用这种方法。有损和无损编码都由相同的结构或软件实现。这被称为可逆变换，并且基于无损模式。有损模式是通过切割码流来实现的。考虑到提升步长的传递函数，每个子带上基函数的长度是可变的。通过在合成块中加长低通滤波器的长度，减少了在DCT中引起的块效应。通过缩短带通滤波器和高通滤波器的长度，还可以减少振铃伪影。计算机仿真表明，该方法在有损模式下与传统方法兼容，在无损模式下有时比传统方法好得多。

We proposed two method for lossy and lossless image coding. One is based on wavelet transform which is implemented by two-channel IIR filter banks consisted of the parallel connection of allpass filters. This structure is very simple and less computational complexity, but the parameters for each mode are individual each other to get good performance. Another is based on lapped biorthogonal transform which is consisted of integer DCT and lifting steps to overcome the problem as mentioned above. By this method. Both lossy and lossless codings are implemented by the same structure or software. This is called reversible transform and is based on lossless mode. Lossy mode is realized by cutting bit stream. Taking consideration with the transfer function of lifting step, the length of bases function on each subband is variable. The blocking artifacts of which causes in DCT are reduced by making the length of lowpass filter long in the synthesis bank. Ringing artifacts are also reduced by making the length of bandpass and high pass filters short. Computer simulations show that the proposed method is compatible with the conventiona method in lossy mode and is sometimes much better than the conventional method in lossless mode.