基于总变差的图像放大和增强方法

Total Variation Based Image Enlargement and Enhancement

利用小波的多分辨分析和总变差极小原理,提出了一种实现图像放大和增强的新方法。该方法把原图像作为放大图像的小波子带,对放大图像强加了一种小波系数型的限制。放大图像利用图像总变差极小模型进行正则化。经求变分产生带限制的非线性扩散方程作为总变差极小的必要条件,求解偏微分方程得到增强的放大图像。对人工合成图像、医学MRI心脏切片和人物图像进行了实验。实验结果说明该方法同时实现图像放大和增强的有效性。

The authors present a novel variational approach for image enlargement and enhancement.It is based upon that the original image can be deduced from enlarged image by a wavelet sub-sampling.This impose some constraints on enlarged image.The authors use the total variation of the image as measure of regularity.The Euler-Lagrange equa-tions with constraints that are deduced from the variational principle by the minimize of total variation.Image is enlarged and enhanced simultaneously.Some experimental results demonstrate the validity of the new method for image enlarge-ment and enhancement.