| Summary: | 碩士 === 國立高雄應用科技大學 === 資訊工程系 === 100 === Images are often affected by different noise interference through the transmission process. The noise causes error information in image analysis or brings quality of images. To reduce the influence of noise, a simple and easy method is to receive a fewer images and averages those images to get a clean image, but this method costs huge time. And if it is difficult to obtain the image, the method is more infeasible to implement.
Noise always exists anywhere, for example ISO noise, impulse noise, Gaussian noise and so on. Almost all of proposed methods used the specific filter to conquer the specific noise, however, there are more than one kind of noises existing in the image, i.e., the images are always mixed noises. Here we discuss how to handle the mixed noise in images for general. The median filter has better performance for impulse noise and the average filter is good at lower Gaussian noise respectively. In this thesis, proposed method is based on basic noise filters to design a better filter.
This paper presents a method based on median filter to avoid extreme value and designs an adaptive mask according to the number of extreme values in original mask. Then, finding the similarity for each pixel in the mask and giving weights to each pixel in the mask by the principle of proportionality. To do so, the presented method can achieve the goal which can remove high-intensity impulse noise and reduce mixed noised interference.
|
|---|
