多权值的小波图像放大方法

Multi-Weight Image Zooming Method Based on Wavelet

针对图像放大过程中出现边缘模糊和受噪声干扰的问题,结合各向异性扩散和小波变换的特点,提出了一种新的多权值的小波图像放大方法.对噪声图像进行小波变换,得到高频小波系数.利用多权值的各向异性扩散方法作用于高频小波系数上,实现了对高频小波系数的有效估计.为了保持放大图像与原图像的相似标准,把原图像作为放大图像的低频部分,通过与扩散后的高频小波系数重构得到最终的放大图像.实验表明,该方法易于得到既去除噪声又保持了边缘细节信息的高频小波系数,使得放大图像在纹理和边缘部分都得到了明显的增强,视觉效果也有较大提高.

For the sake of edge blur and the noise problem in the process of image zooming, more weights image zooming method of wavelet is proposed according to characteristic of anisotropic diffusion and wavelet transform. The noise image is decomposed by wavelet transform so as to get high frequency wavelet coefficients. By using the more weights anisotropic diffusion method upon the high frequency wavelet coefficients, the effective estimation of the high frequency wavelet coefficients is got. In order to maintain a similar standard between the zoomed image and the original image, the original image is regarded as the low frequency part of the zoomed image. The low frequency part can be used to reconstruct with the high frequency wavelet coefficients which are obtained by diffusion, then the final zoomed image is got. The experiments show that the new method is easy to get high frequency wavelet coefficients which do well in resisting image noises and keeping the detailed edges. It makes zoomed image significantly enhance in the texture and the edge portions. The visual effects are improved greatly.