一种改进的全变分自适应图像去噪模型  

An improved adaptive image denoising model based on total variation

作　　者：侯榆青[1] 张欢[1] 史晶[1] 张玲艳[1] 

机构地区：[1]西北大学信息科学与技术学院,陕西西安710127

出　　处：《西北大学学报（自然科学版）》2008年第3期371-373,共3页Journal of Northwest University（Natural Science Edition）

基　　金：国家自然科学基金资助项目(10671156)

摘　　要：目的比较几种不同的变分图像去噪模型的优缺点,提出一种新的根据局部梯度信息自适应调整光滑性测度的模型。方法新模型增加了梯度模值与两个门限的比较,小于门限1采用各向同性扩散去噪模型,大于门限2采用TV去噪模型,梯度模值在两个门限之间的采用自适应去噪模型。结果新模型在强噪声水平下,改善了人眼主观视觉感受,均方误差(MSE)降低了约35,峰值信噪比(PSNR)提高了约2.25dB。结论新模型改进了原有模型,进一步减少了"阶梯"效应,得到了更好的去噪效果。Aim To compare the advantages and disadvantages of some models for image denoise based on total variation and propose a new adaptive model which can adjust its smooth measurement according to local gradient.Methods The comparisons between the gradient mold and two thresholds are introduced in new model.The isotropic diffusion denoising model is used if the gradient mold is less than threshold 1,and TV denoising model is used if the gradient mold is greater than the threshold 2,and adaptive denoising model is used if the gradient mold is between two thresholds.Results Human subjective vision feelings are improved,and the mean square error（MSE） reduces by 35 and the peak signal-to-noise ratio（PSNR） increases by 2.25dB under the strong noise level.Conclusion The original model is improved.＂Staircase＂ effect is reduced further more,and better denoising effects is got by new model.

关 键 词：图像去噪 全变分 自适应 梯度模值 门限 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]