Summary: | 碩士 === 國立宜蘭大學 === 電子工程學系碩士班 === 107 === Noise suppressing is a necessary process for image and video. Today, many excellent image filter have been provided and applied widely. However, there is few discussion of processing Random-Value noise. Especially, literature about video noise processing with random value impulse noise is fewer. This thesis is mainly based on Peer-Group image filter which is proposed previously by laboratory. It seems to have good effect after test, but not that preferable when applying to processing video. This thesis would improve this method and apply it to videos. The proposed method mainly consists of three parts – detection, filtering and secondary verification. We first covert detected pixels to fuzzy value according to fuzzy theory, then use the concept of Peer-Group with directional, double subsidies directional and similarity module to detect and filter random noises. At last, because the settings and threshold of previous detection and reduction are roughly, we conduct secondary verification on the pixels which are justified as noise-free. This could diminish some misjudgements. The proposed method basically uses addition and subtraction operations to detect and reduction filter and avoids complex operation. Simulations confirm that the novel filter has significant improvement on detection and reduction when applying this method to video.
|