An Arithmetic Coding Scheme for Gray-Scale Images with Minimum Description Length Criterion.

In the arithmetic coding of the multilevel gray-scale image (128 levels) up to now, the fixed model for the gray-scale image is supplied. This paper proposes a method to improve the compression ratio by introducing the minimum description length (MDL) into the encoder to describe more precisely the local property of the image and by adaptively selecting the probabilistic model. In the course of this development, a practical form of MDL criterion is presented to be adaptively applied to the encoding as an approximation to the MDL criterion proposed by Rissanen. A problem in the arithmetic coding of the multilevel gray-scale image using a probabilistic model with a large number of parameters is the deterioration of the compression ratio. To solve this problem, this paper proposes a convergence acceleration method for the average code length for the arithmetic encoding of the high-level gray-scale image. Then, based on the proposed acceleration method, the MDL criterion is introduced into the arithmetic coding of the multilevel gray-scale image. Finally, the effectiveness of the proposed coding method is demonstrated by computer simulation. [ABSTRACT FROM AUTHOR]