一种基于快速傅里叶变换的图像条带噪声去除改进方法

An Improved Method for Image Stripe Noise Removal Based on Fast Fourier Trasform

摘要:

传统的基于快速傅里叶变换(Fast Fourier Trasform,FFT)的图像条带噪声去除方法,即使用指数低通滤波方法使图像的高频分量被滤除,会造成图像细节信息的丢失。文章提出了一种基于 FFT的条带噪声去除改进方法,即空间域插值与频域滤波相结合的条带噪声去除方法。该方法利用空间域去噪后的高频分量替换传统频域滤波频谱图中的高频信息,避免了由于地物某些高频信息与条带噪声频谱混在一起而造成的高频信息丢失。采用信噪比(Signal to Noise Ratio,SNR)和峰值信噪比(Peak Signal to Noise Ratio,PSNR)两种评价指标分别对传统的指数低通滤波方法和改进的滤波方法的去噪效果进行评估。基于改进方法处理后的图像,其 SNR和 PSNR均有明显地提高, SNR值最大可由 27.12提高至 49.87,PSNR值最大可由 37.89提高至 61.00,表明该方法能够有效去除条带噪声。

Abstract:

The traditional method of image stripe noise removal based on fast Fourier transform (FFT), that is, exponen-tial low-pass filtering, causes the high-frequency component of the image to be filtered out, resulting to the loss of image details. An enhanced strip noise removal method based on FFT is proposed, which involves a combination of spatial do-main interpolation and frequency domain filtering. In this method, high-frequency information in the traditional frequency domain is replaced by high-frequency components upon spatial domain denoising, which prevents the loss of high-frequency information caused by the mixing of certain ground object high-frequency information with the band noise spectrum. Two evaluation indexes, signal-to-noise ratio (SNR) and peak-to-noise ratio (PSNR), are employed to evaluate the denoising effect of the traditional exponential low-pass filter and the enhanced filter respectively. Additionally, the im-age processing based on the improved method yields a favorable effect, with significant improvements observed in SNR and PSNR. The maximum value of SNR increase from 27.12 to 49.87 and the maximum value of PSNR increase from 37.89 to 61.00. It suggests that the method effectively eliminates strip noise.