An Approach Of Adaptive Median Filtering For Color Image

Color image signal is often interfered by many kinds of noise when it is produced, transmitted and recorded. Filtering is to weaken the interruption of the noise to the image and improve the quality of the image. And what is more: it would help edge detection, image segmentation, feature extraction and so on. Generally, a good filter would contain a wonderful filtering of noise and keep most details of the image. And it is different for color image from normal image for filtering.In this article we concluded some specials of color image, besides we had improved the process of median filtering, and get to some result as below:1. We studied the detection of noise in the color image, which is very important in the process of filtering. Different window would give different result of the detection of noise. So we should choose different threshold in the filtering.2. Studied the traditional way of median filtering of color image, which gave the premise of new process.3. According to the RGB color space, a new way to remove the impulse noises in an image is proposed. Firstly, in each color channel of an image, the locations of impulse noises are detected in serial mode. With the judgment threshold adjusted adaptively according to the average value of the four fields surrounding the pixel, the pixels of impulse noise are detected. And, the detected results of impulse noises of every color channel image are synthesized to the final result of the whole color image. On the basis of the results, an approach of improved median filtering is applied to selectively remove the impulse noises in a color image. Simulation experiments showed that the approach, which is integrated by both modes of scalar quantity and vector, can remove the impulse noises of color image effectively and have better performances than the other existing approaches of color image filtering.