CUDA_Median_Filering

: The Lab exercise is to design and implement the Median Filter at CPU and GPU. Median Filter is useful filter in computer vision for removing salt and pepper error. The lab give exposure to CPU implementation of median filter using Bubble sort algorithm. Also, Filter implementation is being done on GPU with and without shared memory. The performance comparison is being done for image recreation. Three different size images are being used to get the output and compare results for median filtering. Leena.bmp is of size “252 x 205”, RIT.bmp is “946 x 532” and milkyway.bmp is “4000 x 2000” pixels. The performance and speedup for GPU implementation of the filter is being analysed. Finally the results are represented in a table showing timings and speedup. The implementation for median filter is good example to implement using shared memory in GPU and analyse the working and behaviour of GPU.