Image Filtering Algorithm And Application In The Test System For Electromagnet's Performance

With the rapid development of digital products, image processing is used in a lot of fields. During acquisition and transmission in practical application, images are often subject to several types of noise, which will make the image quality deterioration; therefore, image denoising is very important before the processing. Image denoising filtering algorithm directly affects the results, so research on the filtering algorithms is very practical.Image filtering algorithms usually from both spatial and frequency domain to study, spatial filtering algorithms are based on the template window image space by the pixel of the neighborhood to operate, the process is relatively simple, using a small filter templates denoising effect is obvious. Frequency domain filtering algorithms are the frequency variable in the Fourier transform and operate within a defined space, which make the frequency's change and intensity of gray correspond, the low frequency components are image's background and overall profile, the high frequency components are image's edges and noise. It often uses this feature to remove noise.In this paper, it has simulation of the spatial filtering algorithms and the wavelet transform algorithms, introducing neighborhood averaging filter and several improvement median filters which are presented this year, the above methods have some shortcomings, so this paper presents an improved median filter. The improved median filter uses a sliding big window block to accelerate the processing speed, which can make the system real-time, In order to avoid image blurring, it adds the conditions to distinguish noises and edge detail, and then use neighborhood averaging filter to remove image border's noise points. The simulation results show that the algorithm is easy to implement and the effect is good, especially for the high density impulse noise. Wavelet transform is to introduce the threshold filtering algorithms, it presents two improved algorithms of continuity and adaptive, which are based on hard threshold. The soft (hyperbolic) threshold are used in the threshold value range after setting two thresholds, and then remove the isolated points according to the adjacent correlation coefficient during the process. As a result, the improved algorithms are better than hard threshold, it shows that the improved algorithms have both better visual effect and PSNR than the traditional approaches after simulate. At last it descript the test system for electromagnet, and during the curve test, we find the sensor jitter cause noise in the curve information, we have the image denoising algorithms be used in the system because of the actual characteristics of the noise, and then verify the algorithms in Matlab 7.0.