基于多模板规则的多元映射图象平滑方法研究

基于超越函数和多分辨分析的思想，提出了一种图象空域平滑方法。该算法依据图象多层邻域卷积的一致性，自适应地选择平滑单元。它具有保持图象细节、稳健性较强和运算速度较快等优点；还提出了衡量图象清晰度的局部误差。

A Multiple Unit Transform for Image Smoothing Based on Two MultipleMask Rules

A common problem in image processing is the difficulty caused by two conflicting requirements: (1) suppression of random noise, and (2) minimization of the blurring of the edges of image. In this paper, we present what we call multipleunit smoothing method, which is based on multipleunit transforms with two multiplemask rules. The smoothing algorithm proposed can provide a number of distinct advantages over spatial domain approaches. The most important one is that the edges in an image may be fairly accuratedly preserved while retaining robust suppression of random noise in the flat regions of the image. This is accomplished through a parallel multipleunit transform using a couple of linear functions and associated rules. Our algorithm is compared with meanfilter and medianfilter smoothing algorithms in Table 1 (for blurring) and Table 2 (for real value). In both Table 1 and Table 2, our algorithm gives the largest reduction in noise as the lastrow values for MAE (mean absolute error) and MSE (square foot of meansquare error) are both the smallest Also, in both Table 1 and Table 2, our algorithm gives the least blurring of image edges as the lastrow values for PMAE (partial mean absolute error) and PMSE (square root of partial meansquare error) are the greatest.