摘要
在医学图像的显示、处理和分析中 ,常常需要将图像进行插值处理。本文在PC机上实现Sinc插值、最近点插值、线性插值、三次插值、B样条近似以及拉格朗日多项式插值等Scene based方法 ,并将其应用于医学图像处理插值的三个主要领域 ,然后用相关系数、均方差、差异点比例、相对时间等统计量对各种方法进行了评估比较。为了便于观察使用图像相减方法对插值引入的误差进行了显示。结果说明三次多项式插值和Sinc cos插值可以获得较好的插值效果 ,在使用较小的插值算子时 ,可考虑使用 2× 2的Cubic插值来代替目前最常用的线性插值。
The technique of image interpolation is usually required for medical image processing, analysis and display. Programs were performed on PC to simulate the three main situations (resolution change, image rotation, view angle change) typically required in medical applications. In this paper, results obtained by several scene-based interpolation methods, such as truncated and windowed sinc, nearest neighbor, linear, cubic, B-spline and Lagrange polynomial interpolation were compared. In each case, the efficiency and accuracy of a particular interpolation technique were evaluated by analyzing its visual quality, statistical interpolation error and relative runtime. Summarizing the evaluations, cubic and sinc-cosine interpolator provided better results. When using small interpolation operator, it may be considered to use the 2×2 cubic interpolator instead of the linear interpolator most often used currently.
出处
《北京生物医学工程》
2002年第1期19-23,共5页
Beijing Biomedical Engineering
