Study On Digital Video Spatio-temporal Noise Reduction Algorithms

With the development of computer communication network and multimedia technology, digital video technology is widely used in HDTV, video conferences, medical equipment, education, and civil or military fields, because it is convenient in storage, transmission, switching, coding ,and it is more stable, reliable and interactive than the analog video technology. The quality of digital video will affect the video products directly. Due to the inherence of electronic products, outside interferences and other factors, it will introduce noise into the digital video productions and transmission links. The noise not only makes the image's quality degraded, but also affects the following processes, such as image coding, decoding, analysis and processing. So the reduction algorithms are required.The difference between video image noise reduction and single image noise reduction is: video image is a group of single image sequence with continuous contents, which has more reference information than a single image. Taking into account that there are correlations between adjacent frames of video sequence, researchers proposed a block-based average time-domain filtering, block-based motion compensation time-domain filtering in video image noise reducing. These processes are less in time-consuming and better in noise reduction, but will introduce the artifacts into next frames because of rotation and scaling.This paper presents a video image noise reduction algorithm which can not only get better result in noise reduction but also avoid artifacts. Firstly, this article describes some video image sequence noise reduction theories Secondly, introduces the digital video image characteristics and common noise reduction approaches. Thirdly, combining with the existed temporal filtering algorithm and a content-based spatio-temporal varying filtering algorithm (STVF), proposes a spatio-temporal adaptive video noise reduction algorithm with the accuracy of motion detection and a bilateral spatio-temporal adaptive filtering. This algorithm could remove the trailing using the temporal filtering after motion estimation, and avoid blurring and block artifacts with spatio-temporal filtering. By experimenting in different standard sequences with different noise pollution, it shows that the algorithm can produce a good video sequence with clear outline and avoid artifacts.