Demonstrate Median Filtering Of An Image

Abstract

Image processing plays an essential role in improving visual quality and extracting meaningful information from digital images, especially when they are degraded by noise during acquisition or transmission. Among various denoising approaches, median filtering is a widely adopted non-linear technique known for its ability to suppress impulse noise, particularly salt-and-pepper noise, while effectively preserving edge details. This study presents a demonstration of median filtering applied to both grayscale and color images contaminated with different noise levels. The filtering process operates by sliding a window across the image and replacing each central pixel with the median value of the neighboring intensities, which helps in smoothing unwanted noise without significantly blurring important structural features. In this work, several test images are intentionally corrupted with salt-and-pepper noise and subsequently restored using median filtering with different window sizes such as 3×3 and 5×5. The performance of the method is evaluated through visual comparison between original, noisy, and filtered images, along with quantitative metrics including Peak Signal-to-Noise Ratio (PSNR) and Mean Squared Error (MSE). The analysis also investigates the influence of window size on noise reduction capability and edge preservation. Experimental results indicate that smaller windows maintain fine details more effectively, whereas larger windows provide stronger noise suppression at the cost of slight detail loss. Overall, the findings demonstrate that median filtering remains a simple, efficient, and reliable technique for removing impulse noise while retaining significant image features, making it suitable for various digital image processing applications.