带噪声彩色图像的改进修复方法

Improved Method for Color Image Inpainting and Denoising Based on Total Variation Model

图像修复是修复图像中破损区域的信息或去除图像中多余的物体的技术。基于全变分模型(TV)的彩色图像修复,由于求能量最小时图像的Euler-Lagrange等式在整个图像区域内有效,尽管在降噪的同时也能进行修复,但不能很好地保持彩色图像的轮廓信息,处理边界比较模糊。故提出对含有噪声彩色图像先进行降噪,得到好的降噪图像,再对降噪后的图像进行修复。即通过判断像素点是否在修复区域内部还是外部来确定λe取值,得到了分别用于彩色图像修复区域外部降噪能量方程的降噪梯度下降流方程,和用于彩色图像修复区域内部修复能量方程的降噪梯度下降流方程,完成修复区外部的图像降噪和修复区内部的图像修复。实验表明,这种方法既能很好的降噪,保持彩色图像的清晰边界,又能取得很好的修复效果。

The goal of digital image inpainting is to restore damaged regions or remove objects in the image. Color image inpainting and denoising based on total variation model can inpaint while denoise, but it cant keep good outline information and makes the borderline more blurry. In order to overcome this shortcoming, this paper presents a novel inpainting and denoising algorithm based on color image total variation model, which first does the work of image denoising; then inpaints the denoised image. This novel algorithm makes image much better and keeos clear borderline. The result show that the. inpainting effect is very well.