Noise Reduction For IP Camera Based On Spatiotemporal Filtering

With the improvement of life, people’s requirement for life quality is becominghigher, especially in visual products. At the same time, the rapid development ofdigital multimedia technology makes digital system more advantageous thansimulation system. Unfortunately, noise is introduced inevitably in the digitalizingprocess. On one hand, the existence of noise seriously weakens the observationquality of digital video image sequence. On the other hand, it brings a lot of pressurefor follow-up video image sequence processing. Therefore, how to remove the noiseof video sequences has always been a hot topic of discussion and researching.In this dissertation, the researches of spatiotemporal denoising algorithm areanalyzed and then advantages and disadvantages of the industry’s common methodsare summarized and compared. Based on the above work, a new algorithmframework of spatiotemporal denoising is designed and realized. From the practicalengineering results and the experimental analysis, this paper has improved thespatiotemporal denoising algorithm. Firstly the motion status of each block isdetected, according to which the denoising solution is selected. And the next step isto detect the edge of image texture and combine the result with the previous motionstatus to determine the integration method. In order to adapt to different applicationscenarios, this paper put forward an improvement of adjusting the intensity ofspatiotemporal denoising algorithm according to the value of current pixel. Inaddition, this paper analyzes the sources and the influence factors of fixed patternnoise in the video sequence through a large number of experiments and proposes thesolution.The solution to decrease fixed pattern noise in the video sequence has beenrealized and verified completely. Now this module has been applied to the product.As for the above improved spatiotemporal denoising algorithm, the hardwareimplementation has been achieved completely. After designing detailed hardwarelogic project, the logic project is developed and realized. Then the simulationexperiment is conducted to verify its performance. Lastly, we build the FPGAplatform to check its actual denoising effect real-timely, achieving better results thanthe last generation product.