A fast noise reduction method based on human visual system

To perform well, an image noise reduction method should be built upon an effective noise detection scheme. However, almost all the noise reduction methods published in recent years are in lack of a well-designed noise detection scheme. As a matter of fact, they are nothing more than only filters for noise reduction, such as the standard median (SM) filter and the center-weighted median (CWM) filter. Although these filters already have a good performance on noise reduction, they will surely be able to do even better if there exists a nice detection method. In this paper, we shall propose a new noise detection method that is based on the human visual effect and is used to reduce the impulse noise of images. The detection function is used to classify each pixel into the group of either corrupted pixels or uncorrupted pixels. After this classification, we shall also propose a filter scheme based on the standard median filter to process the corrupted pixels. In our experiment results, we shall show the performance of our method in comparison with some popular filters in recent years, like the standard median filter and the center-weighted median filter (both without noise detection methods) as well as other filters like the noise adaptive soft-switching median filter, which has a noise detection method different from ours.