一种新的边缘保持局部自适应图像插值算法  

New locally adaptive image interpolation algorithm based on edge preserving

作　　者：孙春凤[1] 袁峰[1] 丁振良[1] 

机构地区：[1]哈尔滨工业大学电气工程及自动化学院,哈尔滨150001

出　　处：《仪器仪表学报》2010年第10期2279-2284,共6页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金(50275040)资助项目

摘　　要：为了尽可能降低噪声的影响和保持图像的边缘信息,提出了一种新的边缘保持局部自适应图像插值算法。算法首先对分割成部分重叠的各子图像进行整数小波变换,然后对于高频子带进行逐点BayesShrink阈值去噪和中值滤波去除高斯和椒盐噪声,同时对于低频子带进行亮度保持的直方图均衡增强图像的对比度,最后根据自适应阈值Canny边缘检测将子图像划分为平坦和边缘区域,分别进行施用双线性插值和非线性插值。实验结果表明,MBW图像插值算法得到的插值图像有较好的主观图像质量,其峰值信噪比(PSNR)比传统方法平均提高了2.23～2.96dB。In order to reduce the influence of noise and preserve edge information, a new locally adaptive image interpolation algorithm based on edge preserving is proposed. First, integer wavelet transform is applied to the partially overlapped sub-images. Then, domain BayesShrink threshold denoising and median filtering are used in the high frequency band to remove Gaussian noise and impulse noise, and brightness preserving histogram equalization is used to enhance the contrast of the image in low frequency band. Finally, the sub-images are segmented into flat areas and edge areas using self-adaptive threshold Canny edge detection, and then bilinear interpolation and nonlinear interpolation are carried out. Experiment results show that, the obtained interpolated image using MBW interpolation algorithm has good subjective image quality, and the average PSNR is improved by 2.23 to 2.96 dB compared with those of traditional interpolation methods.

关 键 词：自适应插值 逐点BayesShrink整数小波 直方图均衡 边缘检测 

分 类 号：TN911.73[电子电信—通信与信息系统]