The Study on Effective Video Noise Reduction Based on Spatial-Temporal Filtering

• Main Authors: Ming-Kun Wu, 吳明坤
• Other Authors: Thou-Ho Chen
• Format: Others
• Language: en_US
• Published: 2005
• Online Access: http://ndltd.ncl.edu.tw/handle/09108338569604417584

碩士 === 國立高雄應用科技大學 === 電子與資訊工程研究所碩士班 === 93 === Noise can be added to still images or video sequences in various steps such as image acquisition, recording, and transmission. And then the results of post-processing tasks are influenced by noise. As a result, noise reduction is important for video processing. For video-noise filtering, it may cause object-overlapped phenomenon in an image frame due to the occlusion problem when the spatial-filtering is only used, excluding the temporal-filtering. Oppositely, the image may be blurred and even the noise can’t be reduced largely if the temporal-filtering is only performed but spatial characteristics of the image are not utilized. Therefore, for removing significantly noise, a video-noise filter should be able to work on both temporal and spatial domains. To improve problems of the object-overlapped phenomenon and blurred edges existed in the previous methods, the thesis presents a motion-compensated spatiotemporal filter based on human visual perception. For filtering on the spatial domain, the just noticeable difference（JND）is used to detect whether the pixel is located on the edge and whether disturbed or not. For filtering on the temporal domain, pre-frames and post-frames are used to estimate for the original pixel and hence the temporal noise can be removed significantly. Based on the above filtering, the video can have a high-quality display according to the human visual perception and further improve the followed compression and segmentation processes.