Median Filter Performance Based on Different Window Sizes for Salt and Pepper Noise Removal in Gray and RGB Images

Autor: Zahra T. Alser, Rasha Eltayeb Abd Elatif, Elmustafa Sayed Ali Ahmed
Rok vydání: 2015
Předmět:
Zdroj: International Journal of Signal Processing, Image Processing and Pattern Recognition. 8:343-352
ISSN: 2005-4254
DOI: 10.14257/ijsip.2015.8.10.34
Popis: Noised image is a major problem of digital image systems when images transferred between most electronic communications devices. Noise occurs in digital images due to transmit the image through the internet or maybe due to error generated by noisy sensors. Other types of errors are related to the communication system itself, since image needed to be transferred from analogue to digital and vice versa, also to be transmitted in most of the communication systems. Impulse noise added to image signal due to process of converting image signal or error from communication channel. The noise added to the original image by changes the intensity of some pixels while other remain unchanged. Salt-and-pepper noise is one of the impulse noises, to remove it a simplest way used by windowing the noisy image with a conventional median filter. Median filters are the most popular nonlinear filters extensively applied to eliminate saltand-pepper noise. This paper evaluates the performance of median filter based on the effective median per window by using different window sizes and cascaded median filters. The performance of the proposed idea has been evaluated in MATLAB simulations on a gray and RGB images. The experimental results show that median filter has a good performance in gray and RGB images in low noise densities and also in high noise densities when using cascaded median filter and high level of window sizes, but with higher window size a degree of blurring effect will be added to filtered noise.
Databáze: OpenAIRE