Fast and Effective Median Filter Algorithms for Salt and Pepper Noise

Abstract

For large data of the image processing in the fields of medication or communication, the salt and pepper noise is one of the key problems to affect the quality of the images. In order to obtain stable, fast, and effective filtering algorithms for salt and pepper noise, this paper proposes four filter algorithms, named fast and effective median filter1 (FEMF1), fast and effective median filter2(FEMF2), fast and effective median filter3(FEMF3), and fast and effective median filter4(FEMF4), respectively. FEMF1 uses a second-order differential method to convert the image to a binary matrix to determine the exact location of the broken pixels. It is stable and fast, however it has a sense of noise. FEMF2 and FEMF3 use different directions to search with changing windows. FEMF4 grows square bar window with unlimited growth and has search functions in all directions. FEMF4 costs the lowest time and presents the high quality performance. The running time of the FEMF2 is only slower than that of the FEMF4. Experiment show that FEMF2, FMEF3 and FEMF4 have stable, fast and effective features for low and high noises. These algorithms can be used in engineering applications.