This project mainly discusses about the design of switching median filter using Very Large Scale Integration architecture for removal of Salt and pepper noise in an image.
Any imaging system's images suffer from arbitrary volatility in intensity values, sudden changes in illumination, and low contrast. Such images are devoid of usable information and are difficult to interpret visually. Subsidiary information from such photos is recovered by applying various filtering algorithms to remove noise, sharpen contrast, and detect edges. One such nonlinear method is median filtering, which removes range isolated noise such as salt and pepper noise while keeping the edge information. When the image is distorted by excessively high impulse noise, however, median filtering fails. In the event of high-intensity impulse noise, switching-based median filtering can remove the noise. This project proposes a VLSI design for a new switching-based median filter to eliminate high-density salt and pepper noise from digital photos. To determine whether a pixel is noisy or not, the absolute difference between the center pixel and the array median generated from a 3 x 3 sliding window is compared to a predetermined threshold value. The noisy pixels in the 3 x 3 filtering window are replaced by the median of noise free pixels during the filtering stage. If a pixel value is corrupted, the median of the next window is used to replace it. The real pixel intensity value is preserved if it is not damaged. Using a threshold detector, we can determine whether a pixel is corrupted or not. We can either feed salt and pepper noise straight into an image or produce it ourselves. This is a stage in picture preprocessing. Using MATLAB's built-in tools, we've imported the image and turned it to a salt and pepper noise image. In our project, we mostly employed three tools. We utilized MATLAB to preprocess the image and plot values to rebuild the original image, then MODELSIM to create the median filter and XILINX to compare the existing bubble sort with the suggested three cell sorter. Simulation results show that the proposed algorithm outperforms traditional median-based filters, and that it is especially useful in cases where the images are severely corrupted.