多种搜索算法在医学图片彩色迁移上的应用与分析

通过对真彩图像与灰色MRI(核磁共振图像)切片在亮度及纹理等特征上进行配准而实现灰度MRI切片的彩色化,可以得到有效的符合真实人体器官组织的真彩仿真图片。对多种搜索算法如穷举法、随机法、粒子群算法和遗传算法与点邻域亮度分布纹理和局部颜色结构分布相结合并应用于Welsh图像彩色化的算法进行研究和分析。着重分析和修正遗传算法在Welsh彩色化上的应用。该算法将源图像块和目标图像块的亮度特征以及纹理特征构成适应度函数,并依据代数的增加而降低交叉和变异的概率以减少进化后期的跳跃性。其搜索的彩色特征种群经过选择、交叉和变异等操作后逐代进化而得到最终彩色化图像点。最终彩色化的MRI切片的三维重构模型能多层次清晰地反映器官组织的分布和构造。

Application and Analysis of Multi-searching Algorithms to Color Transter between Medical Images

This paper researches transferring a gray MRI slice into a colorful one that is valid and matches the human organs through registering the features of brightness and textures between the true colorful and the gray images.This paper probes a Welsh color transfer algorithm based on the combination of several searching methods,and the features of brightness within the neighborhood of a pixel and local color distribution.It particularly analyzes and modifies GA to adapt to the Welsh color transfer algorithm.This method will use the brightness and texture features of source and target images as the fitness function,and decrease the probability of crossover and mutation when the population evolves to avoid the mutation of the fitness of the population in later stage of evolution.The population which is used for searching by color features evolves through selection,crossover and mutation to get the fittest individual.3D color model reconstructed from the colorized slices can clearly reveal multi-level structure and distribution of human organs.