基于Curvelet变换的阈值补偿图像去噪方法研究

ON IMAGE DENOISING APPROACH OF THRESHOLD COMPENSATION BASED ON CURVELET TRANSFORM

基于Curvelet变换硬阈值图像去噪方法,能有效地抑制图像中的噪声。为了减小在去噪过程中产生的伪吉布斯现象,提出一种阈值补偿的去噪方法,即用软阈值和"过扼杀"系数一起补偿硬阈值。试验结果表明,该方法在去噪和保持边缘的同时,取得了较好的视觉效果,并且PSNR也得到一定的提高。在同等噪声条件下对比不同的噪声图像,与硬阈值去噪算法相比,阈值补偿算法的峰值信噪比提升为0.44%-0.96%;同一图像在不同的噪声条件下,与硬阈值去噪算法相比,其峰值信噪比提升了0.44%-0.76%。

Hard threshold image denoising method which is based on the Curvelet transform can effectively suppress noises in image. To reduce the psudo-Gibbs phenomenon when denoising, a noise removal method named threshold compensation is presented, that is, to compensate the hard threshold with the soft threshold and the ＂over-killing＂ coefficient together. Experiments show that the method can remove noise and remain edge, while the distortions are suppressed efficiently with good visual effect, and PSNR is also enhanced to certain extent. Comparing with the hard threshold denoising algorithm, the PSNR of the threshold compensation algorithm is enhanced up to 0.44% -0.96% when contrasting different noise images in condition of same noise level, and up to 0.44% -0.76% when contrasting the same noise images in the condition of different noise level.