基于边缘定向增强扩散模型的图像去噪

针对基于全变分(TV)的图像去噪模型,恢复图像存在阶梯现象(staircase)的缺点,提出了一个新的图像恢复的变分模型.定义了一个新的包含图像的边缘位置和方向信息的能量泛函,使得沿图像边缘的切线方向具有较强的平滑能力,而法线方向平滑较弱,可以较好的定位边缘.并且该模型能增强阶跃性边缘,防止因平滑造成的边缘模糊现象.试验结果表明:该模型可以部分的解决基于TV的变分模型中出现的阶梯现象,并且具有较好的去噪效果,而且它还能增强图像边缘并保持边缘的位置,其峰值信噪比(PSNR)在高噪声水平下,较其它方法提高大约1.0dB左右.

Image Denoising Based on Edge-directed Enhancing Diffusion Model

Directing at the drawbacks of the image edge may be smoothed in the image denoising of TV model,an edge-directed enhancing diffusion model is proposed,which adds more information of image edge’s location and direction in the energy function of image,so the edge of the tangential direction has strong ability to smooth,and gradient direction smoothing is weak so it can get right edge.And the model can enhance the step edge to prevent fuzzy phenomenon caused by the smoothing.The experimental result shows the new model has better denoising results than the known diffusion model,and achieves at least 1.0dB gain over other variational denoising methods for peak signal-noise ratio(PSNR) measurement.