摘要
为提高在图像降噪过程中对图像细节信息的保护能力,提出一种基于双树小波和神经网络的图像降噪与增强算法。通过Canny算子检测图像的边缘,通过shearlet变换将噪声图像分解为高频子带和低频子带;使用卷积网络保留边缘区域,通过两层剪切波滤波器组对非边缘区域进行降噪,通过神经网络对总体图像进行增强。实验结果表明,该算法可以实现较高的降噪性能,有效地提高图像的质量。
To improve the protection capability for image detail information during the image denoising process,a double tree wavelet and neural networks based image enhancement and denoising algorithm was proposed.Canny operator was adopted to detect the edges of the image,and noisy image was decomposed to high frequent sub-band and low frequent sub-band though shearlet transform.The convolution neural networks were used to preserve edge regions,two-layered shearlet transform filter banks were used to denoise the other regions,and a neural network was further used to enhance the total image.Experimental results show that the proposed algorithm can realize good denoising performances and improve the image quality effectively.
作者
刘文辉
许瑞
LIU Wen-hui;XU Rui(College of Information Science and Technology,Xinjiang Education Institute,Urumqi 830043,China)
出处
《计算机工程与设计》
北大核心
2021年第5期1402-1408,共7页
Computer Engineering and Design
基金
新疆维吾尔自治区重点实验室专项基金项目(2019D04024)。
关键词
图像降噪
图像增强
卷积神经网络
人工神经网络
小波滤波器组
剪切小波变换
image denoising
image enhancement
convolutional neural networks
artificial neural network
wavelet filter banks
shearlet transform
