摘要
受传统算法的启发,根据图像相邻像素的关联性,扩充图像边缘像素,采取首行与首列(或者末行与末列)填充的方法,利用ERBF核函数做已知像素点的曲线拟合,解决了采样点中不包含原始图像像素点的缺陷。试验表明,基于ERBF核函数和边界填充的图像插值算法,对原始图像进行隔行隔列抽取25%的像素点得到的低分辨率图像,插值后生成的高分辨率图像和原始图像相比,性能参数PSNR高达33.81dB,高于目前的插值算法,没有阈值选择的困扰。而且,该文提出了用差值图AMI来分析不同算法插值生成图像的比较方法。
Based on the relevancy among the pixels in the vicinity, expanding the pixels on the edge of the image, filling the expanding areas with the first row and column, making curve fit from known pixels by ERBF kernel function, it solves the problem dexterously of sampling pixels, that do not include the original image pixels. Experimentations suggest that, when the low resolving image, obtained by taking out 25% pixels with even rows and even columns from original image, become the high resolving image after interpolating, the parameter PSNR of the latter is 38.53dB, which is higher than any other image interpolation algorithms. And, there is no need to choose a threshold, which is always a difficult problem. It also puts forward a new effective method to compare different image interpolation algorithms by absolutely minus image for the first time,
出处
《计算机工程》
CAS
CSCD
北大核心
2007年第2期160-162,共3页
Computer Engineering
基金
国家自然科学基金资助项目(60075019)
关键词
ERBF核函数
边界填充
图像插值
像素拟合
差值图
ERBF kernel function
Boundary filling
Image interpolation
Pixel fit
Absolutely minus image
