Mtf-based Remote Sensing Image Restoration Technology

The remote sensing images are captured by the satellite on orbit, In some cases, we may get the blurred remote sensing images. There are many causes, such as the atmosphere, the remote sensors, and the signal transmission. So it's important to improve the quality of remote sensing images.There many different ways of image restoration. However, lots of these ways are based on the PSF, or need other conditions of imaging system. Sometimes, we can't get enough conditions, and it is difficult to calculate. In these cases, it is hard to restore the images by the usual methods. So, in this paper, we will discuss a different method of image restoration, which does not need the knowledge of the PSF or other conditions.In this paper, we did some research as follows:1. Firstly, we removed the noise of the images. Aim at different noises, we used different methods. For the strip noise, the frequency method based on Fourier Transform was used, and local self-adaptation Median Filter was adopted for the isolated noise. These two methods could remove the noise effectively, at the same time, prevent the images from blurring.2. Then, we restored the images. The blurred images could be improved based on the degradation characteristic of MTF in frequency area. First, the degradation model was established based on MTF. Second, transformed the images from spatial area to frequency area by Fourier Transform. And then, restored the images based on the degradation model in frequency area. At last, transformed the images to spatial area. By this way, we could improve the images in a degree.3. Also, this paper discussed the relations between MTF and PSF, and established PSF model based on MTF. Different types of PSF, including 1-D and 2-D PSF, were all discussed in the experiment in order to acquire the suitable PSF model. And we adopted this PSF model in some usual image restoration algorithms. The results proved that this method can get better images.The methods in this paper are suitable not only for remote sensing images, but also for other kinds of images. However there is still some improvement to make in future research.