小波域的纹理图像快速修复算法

Wavelet-domain fast inpainting algorithm for texture image

针对传统的空间域纹理图像修复算法计算量大、修复时间长的缺点,本文提出了一种纹理图像的快速修复算法。该算法的基本思想是在小波域中利用小波系数的能量来确定待修复块的填充顺序,并结合纹理合成的方法填充待修复区。实验结果表明,该算法不仅可以大大提高纹理图像的修复速度,而且在峰值信噪比和主观视觉效果上都优于传统的图像修复算法。

Traditional spatial-domain inpainting algorithms for texture image need complicated computation and consume a long time. A fast inpainting algorithm for texture image is proposed by determining the order of the filling blocks using the energies of wavelet coefficients in wavelet domain, and filling the inpainting blocks using texture synthesis method. Experimental results show that the algorithm not only improves inpainting speed greatly, but also achieves better peak signal noise ratio （PSNR） and higher visual quality.