【算法】Fractional Multiscale Processing

Fractional Multiscale Processing（FMP），分数多尺度处理

为了较好地渲染一幅图像中比较难处理的区域，譬如亮度突变的区域，我们应用了一种新的数学算法——分数多尺度处理。算法的基本思想是，一个滤波器分解为多个基滤波器，基滤波器独立对图像进行处理，互不干扰。这样做的好处是，我们可以更自然地描述灰度的变化，避免出现伪影。FMP无需经过窗位调节来增强图像的可视效果。

Next generation technology 

Fractional Multiscale Reconstruction (FMR): the new mathematical substructure of MUSICA, which further decomposes image components in microscopic fractions for separate processing.Detail is enhanced with pixel-level precision achieved by splitting the multi-scale filter kernels into smaller fractions and by involving all pair-wise pixel differences within the filter kernels. FMR results in a more accurate multi-scale enhancement model, a balanced participation of all filter kernel pixels in the enhancement process, and better preservation of low contrast details next to high contrast steps.

应该是这样吧：传统的MUSICA的架构是图像金字塔，以滤波核作为处理单元的，而FMP是以每个像素作为处理单元，从而做得更加细致。也可以理解为，图像金字塔体系的进化，传统的图像金字塔会有模糊效果，这对细节（边缘）是不利的，FMP就是为了应对这一点。

参考：

Musica hits the high notes in imaging

Fractional Partial Differential Equation: Fractional Total Variation and Fractional Steepest Descent Approach-Based Multiscale Denoising Model for Texture Image

Efficient CT Metal Artifact Reduction Based on Fractional-Order Curvature Diffusion

Some Fractional and Multifractional Gaussian Processes: A Brief Introduction

This paper gives a brief introduction to some important fractional and multifractional Gaussian processes commonly used in modelling natural phenomena and man-made systems. The processes include fractional Brownian motion (both standard and the Riemann-Liouville type), multifractional Brownian motion, fractional and multifractional Ornstein-Uhlenbeck processes, fractional and mutifractional Reisz-Bessel motion. Possible applications of these processes are briefly mentioned.