Study on A Highly Efficient Loss-Less Coding Scheme for Image Signals

本文提出了一种高效的无损图像编码方案。图像编码基本上分为两个块。一个是模型部分，另一个是熵编码器。首先，我们提出了一种三维马尔可夫模型编码方案作为无损图像编码的模型部分。在我们的模型中，利用了位平面之间的相关性。用12个参考点的三维马尔可夫模型对256个层次的图像进行了计算机模拟，结果表明，与传统的1个参考点的马尔可夫模型相比，该方法最大可使图像的熵降低8%。接下来，我们提出了两种熵编码方法。第一个是关于算术编码的。在我们的编码方案中，通过估计LP的概率为1/(1+2^q)，任何增加的状态都被转移到与符号值无关的相同状态。在该方法中，仅使用准备好的增值表通过移位操作来计算增数数据。使用二值图像，从编码效率和编码操作次数两个角度进行了仿真实验。第二种是可变固定码(VF码)，即高速可译码。我们发现，在一定条件下，VF码比传统的霍夫曼码更有效。通过使用上述模型和熵编码方案，描述了高效的无损图像编码方案。

In this study, a highly efficient loss-less image coding scheme is developed. Image coding is basically separated into two blocks. One is the model part and another is the entropy encoder. First, we proposed a three dimensional Markov model coding scheme as the model part of the loss-less image coding. In ourmodel, correlations between bit-planes are utilized. The computer simulation is performed for the images of 256 levels using the three dimensional Markov model of 12 reference elements and we can get results that this method can reduce the entropy by 8 percent at its maximum compared with the conventional Markov model of 1 reference pixel. Next, we proposed two entropy coding methods. the first one is concerned with the arithmetic coding. In our coding scheme, any augend state is transited into a same state independent of symbol values, by estimating probability of LPS to 1/ (1+2^Q). In this method, augend data are only calculated by shift operation using the prepared augend table. Simulation experiments from points of view in coding efficiency and the number of coding operations are performed using binary images. The second one is concerned with Variable-to-Fixed code (VF code), that is high speed decodable code. We revealed that the VF code is more efficient than the conventional Huffman code under some condition. By using above mentioned model and entropy coding scheme, the highly efficient loss-less image coding scheme is described.