融合分形小波图像去噪和插值的新算法

Joint fractal-wavelet image denoising and interpolation

提出一种能同时进行图像去噪和插值的分形小波算法。该算法在图像的编码过程中实现去噪,在解码过程中完成插值。基于分形的图像去噪算法涉及由含噪图像的统计特性预测无噪图像的分形编码参数,并由这一参数产生含噪图像的无噪估计。通过在解码过程中添加适量的分形小波外推步骤,即可实现图像的插值。为了减小失真、增强经此算法处理后图像的质量,该引入Cycle Spinning算法。实验结果表明,该方法在同时实现图像去噪、压缩和插值上是最优的。且Cycle Spinning算法的引入明显改善图像视觉效果,显著提高图像的PSNR值。

In this paper,a simple and effective fractal-wavelet simultaneous image de-noising and interpolation scheme is proposed.The de-noising is performed during the encoding process while the interpolation is performed during the decoding process. The fraetal-based image denoising involves predicting the fractal code of the original noiseless image from the statistics of the noisy observation.This fraetal code can then be used to generate a fraetal-wavelet denoised estimate of the original image.The fraetal-wavelet interpolation can be easily achieved during the decoding process by performing a suitable number of additional fractal-wavelet extrapolation steps.The cycle spinning algorithm can also be incorporated in the proposed fractal-wavelet joint denoising and resizing scheme in order to reduce some of the artifacts and enhance the visual quality of the denoised and resized estimates.The experiment results indicate that this method is the best in the field of simultaneous image denoising,compression and interpolation.Incorporating cycle spinning algorithm can get better visual effect and high PSNR value.