Performance Analysis of Algorithms over FPGA for Removing Salt and Pepper Noise

This paper shows the possibility to improve the accuracy of medical diagnosis on contaminated radiographic image with salt and pepper noise using FPGA-based filters. Investigation is carried out using filters described widely in the literature, such as Blur, Smooth and Gaussian linear filters and nonlinear median filter implemented on FPGA in an attempt to eliminate the noise. Implementations are checked on a development board with Spartan 6slx100fgg484-3 FPGA and a low-power optimized version Spartan 6slx100lfgg484-1l. The results are verified through the filtered images, resources consumption and the lowest response time. The implementation methodology uses model-based engineering involving MATLAB/Simulink and System Generator software using the XSGlmgLib library to perform co-simulation. The results show that the FPGA implementation of the median filter has a good tradeoff between resource consumption and noise removal efficiency.