All types of filters can be divided into the following classes: frequency, linear, nonlinear, combined. In digital signal processing widely linear filtration methods are used. However, linear filtering demonstrates good results only in the presence of Gaussian additive noise. In the case of impulse noise nonlinear filtering methods, in particular median and rank filtering, work more effectively. In the case of combined noise, linear and nonlinear filters are used sequentially, or a combination of these filters to enhance their positive properties.
Gaussian filtering. The simplest idea to remove noise is to average the pixels in the surrounding area. The simplest option for filtering is to assign a central one pixel of the new value as the arithmetic mean of all its neighbors, the values of which differ from the central value by no more than a certain threshold.
Median filtering. This is the standard way to reduce impulse noise. For each pixel in a certain neighborhood (window) is the median value and is assigned to the pixel that is processed. Determining the median value: If you sort an array of pixels by their value, the median will be the middle element of that array.
Wiener filtering is filtering by minimizing the standard deviation. The method is based on the consideration of images and noise as random processes, and the task is set as follows, to find an estimate of the pixel parameters for the filtered image for which the standard deviation would be minimal.
The analysis of classical filtering methods, such as median filter, Gauss filter, Wiener filter, used in well-known modern image processing software packages, often do not have satisfactory efficiency in processing specific types of noise and often lead to partial loss of useful signal. / p>
Therefore the main task of this work is to develop, program and analyze a new authoring filtering method , which improves the results of data recovery compared to existing approaches, creating a online service image processing under the influence of pulsed noise of different types.