计算机工程与设计
計算機工程與設計
계산궤공정여설계
COMPUTER ENGINEERING AND DESIGN
2015年
8期
2152-2156
,共5页
张芳%张权%崔学英%董婵婵%刘祎%孙未雅%白云蛟%桂志国
張芳%張權%崔學英%董嬋嬋%劉祎%孫未雅%白雲蛟%桂誌國
장방%장권%최학영%동선선%류의%손미아%백운교%계지국
全变差%中值先验%图像重建%小波变换%非局部
全變差%中值先驗%圖像重建%小波變換%非跼部
전변차%중치선험%도상중건%소파변환%비국부
total variation%median prior%image reconstruction%wavelet transform%nonlocal
针对最大后验法只能提供有限的局部先验信息,使重建图像出现阶梯状边缘伪影及过度平滑等问题,提出一种基于小波收缩和非局部的全变差(total variation,TV)中值先验重建算法。在中值先验分布重建算法的目标函数中加入降噪能力优异的TV先验,提出基于TV的中值先验(median prior,MP)重建算法;针对重建图像边缘依然不清晰且存在块状伪影的问题,利用平稳小波变换具有良好的时频局部特性和平移不变性的特点,在基于TV的MP重建算法的每次迭代中,进行小波收缩和非局部降噪,进一步优化图像。仿真结果表明,该算法可以获得高质量的图像。
針對最大後驗法隻能提供有限的跼部先驗信息,使重建圖像齣現階梯狀邊緣偽影及過度平滑等問題,提齣一種基于小波收縮和非跼部的全變差(total variation,TV)中值先驗重建算法。在中值先驗分佈重建算法的目標函數中加入降譟能力優異的TV先驗,提齣基于TV的中值先驗(median prior,MP)重建算法;針對重建圖像邊緣依然不清晰且存在塊狀偽影的問題,利用平穩小波變換具有良好的時頻跼部特性和平移不變性的特點,在基于TV的MP重建算法的每次迭代中,進行小波收縮和非跼部降譟,進一步優化圖像。倣真結果錶明,該算法可以穫得高質量的圖像。
침대최대후험법지능제공유한적국부선험신식,사중건도상출현계제상변연위영급과도평활등문제,제출일충기우소파수축화비국부적전변차(total variation,TV)중치선험중건산법。재중치선험분포중건산법적목표함수중가입강조능력우이적TV선험,제출기우TV적중치선험(median prior,MP)중건산법;침대중건도상변연의연불청석차존재괴상위영적문제,이용평은소파변환구유량호적시빈국부특성화평이불변성적특점,재기우TV적MP중건산법적매차질대중,진행소파수축화비국부강조,진일보우화도상。방진결과표명,해산법가이획득고질량적도상。
Total variation median prior reconstruction algorithm based on wavelet and nonlocal was put forward to solve the prob-lems of the stepladder edge and over-smoothness of reconstructed image resulted from using maximum a posterior.The median prior reconstruction algorithm based on total variation was proposed in which the total variation prior with good ability of noise reduction was added in the obj ective function of the median prior distribution reconstruction algorithm.To make the edge clear and solve the problem of block artifacts of reconstruction images,concerning the stationary wavelet transform having good local time-frequency features and translation invariance,wavelet shrinkage and nonlocal noise reduction were used in each iteration of the MP reconstruction algorithm based on TV so as to optimize the image quality.The simulation results show that this recon-struction algorithm can obtain high quality images.