本論文提出一個基於影像複雜度之脈衝雜訊濾波器。我們經由標準差和受汙染影像的直方圖,區分平滑或複雜區塊。經由模擬結果顯示SWM-I和SWM-II濾波器分別在平滑和複雜區塊有好的效果。因此,我們將SWM-I濾波器和SWM-II濾波器加入雜訊檢測機制,判斷像素是否遭受汙染。根據提出的方法,一張受汙染的影像可以分為平滑和複雜區塊,分別選取適當的濾波器。最終我們經由過濾後影像的峰值信號雜訊比率,驗證雜訊移除能力。 In this thesis, we propose an image complexity based impulse noise filter. We distinguish smooth or complex regions by standard deviation and histogram of the corrupted image. By simulation result, SWM-I and SWM-II filter have better capability in the smooth and complex region, respectively. Therefore, we incorporate the SWM-I and SWM-II filter into a noise detection mechanism to determine whether a pixel is corrupted. According to the proposed method, a corrupted image can be classed as smooth and complex regions the appropriate filters are selected to process the regions, respectively. Finally, we calculate Peak-Signal-to-Noise Ratio of filtered image to perform noise removal capability.