改进各向异性扩散模型在图像滤波去噪中的应用

Application of improved anisotropic diffusion model in image filtering and denoising

为了增强锅炉水位计图像滤波去噪效果,提高图像清晰度,便于后期液位计图像识别研究,通过分析P-M各向异性扩散模型、选择扩散模型及You Yu-Li和Kaveh M四阶偏微分方程的滤波去噪算法,提出了改进各向异性扩散模型滤波算法。所提算法对Perona和Malik两个扩散函数均值化,并引入标准差作为梯度期望值的偏差裕度,结合了P-M各向异性扩散模型保边缘特性的优点,并消除了由于传统各向异性滤波算法迭代过度所造成的阶梯缺陷问题,确保图像有用信息不缺失和像素点平滑度。实验结果表明:所提算法能够更好地降低噪声对目标信号提取产生的影响,提高了图像识别鲁棒性,增强了图像平滑滤波效果,保证了锅炉水位计图像边缘清晰度和完整性。

To enhance image filtering denoising effect for boiler water level gauge and improve image clarity to better study image recognition later, an improved anisotropic diffusion model filtering algorithm is given through analyzing filtering denoising algorithm such as P-M anisotropic diffusion model, selection diffusion model and You Yu-Li and Kaveh M fourth-order partial differential equations. The proposed algorithm makes diffusion functions of Perona and Malik equalization and introduce standard deviation and regards as deviation margin of gradient expectation value, combines with the covering edge feature of the anisotropic diffusion P-M model and eliminates the ladder defect due to the traditional algorithm overshoot so that the image useful information can be kept and pixel smoothness. The experimental result shows that the proposed filtering algorithm can better reduce the impact of noise on target signal extraction and improve image recognition robustness and enhance effect of image smooth filtering, and image edge clarity and integrity of boiler water level gauge.