摘要
由运动模糊图像复原出原图像关键问题是获取点扩展函数和噪声信息,其中噪声通常假设为高斯白噪声。针对匀速直线运动模糊图像,提出一种差分噪声功率估计方法,该方法对模糊图像的差分图进行叠加,放大噪声,并保持每列的噪声互不相关,叠加后的图像的方差与原始噪声方差存在一个数量关系,通过该数量关系来估计出原始噪声方差。实验证明该方法能够相当准确地估计出噪声功率。
The key problem in image restoration is to get the point spread function of motion-blurred images and the knowledge about noise.Gaussian white noise is commonly assumed as the noise.So the power spectrum of noise is important information.A method is proposed to estimate the variance of noise in uniform linear motion blurred images.The method is based on strict mathematical foundation.Experiments prove that the method can accurately estimate the noise power spectrum.
出处
《计算机工程与应用》
CSCD
北大核心
2005年第8期23-25,共3页
Computer Engineering and Applications
基金
国家自然科学基金资助项目(编号:60273064)
关键词
图像复原
噪声
运动模糊图像
image restoration,noise,motion-blurred images
