基于偏微分方程的肉品图像去噪对比研究

COMPARISON OF MEAT IMAGE DENOISING ALGORITHMS BASED ON PARTIAL DIFFERENTIAL EQUATION

肉品中含有水分,或者切割过程中产生的脂肪碎屑或结缔组织,其可见光图像容易产生噪声。基于偏微分方程模型的算法去噪同时能够保持图像的某些特征,常见的有Perona-Malik模型、全变差ROF模型、Y-K四阶模型。通过对Lena图像加入高斯噪声和椒盐噪声,对比了三种模型的信噪比、方法噪声及运算时间。在此基础上,以猪肉图像为去噪对象,比较三种算法的性能。结果表明:ROF模型在去除噪声的同时,保持细节的能力强于其他两种模型,YK四阶偏微分方程模型能够去除噪声,但是图像模糊。去噪效果最差的是P-M模型。

Meat product contains moisture contents as well as fat chips and connective tissues produced during the cutting process,so the noise is common in its image under visible light.The algorithms based on partial differential equation model can remove the noise as well as keep certain features of the image,the widely used models are the Perona-Malik model,the ROF total variation model and the Y-K fourth-order model.In this paper we compared these three models in their signal-to-noise ratios,method noise and operating time by mixing the Gauss noise and the spiced salt noise with the Lena image,and on this basis we selected the pork image as denosing object and compared the performance of these three models.Experimental result showed that the ROF model is superior in maintaining the image details when removing the noise than other two models,the Y-K fourth-order partial differential equation model can remove the noise but with blurring image,and the P-M model is the worst among the three in removing the meat image noise.