抄録
The quantized picture with insufficient quantizing levels often shows spoiling spurious effects such as “false contours” crossing awkwardly through a smooth region in the picture. These effects are usually prohibitive for a realization of the simplest low level quantizing scheme which aims at efficient picture encoding. These spurious phenomena which in a sense are derived artificially from a careless reproduction of encoded picture can however be reduced by realizing a smooth reproduction scheme which carefully assigns continuous but consistent graynesses to every pixel in the picture.
In this paper a new reproduction method of quantized picture with these properties is presented based on the concept of the statistical estimation. Using a Gaussian model of the picture and the quantizer error, the equation of maximum aposteriori probability (MAP) estimation is derived. Although this equation has high dimensions and also is nonlinear, a computer implementation of the equation is achieved by an efficient iterative algorithm based on the direct maximization of posterior density by using the steepest descent method. Several examples of reproduction show usefulness and efficiency of the presented method.
