Research On Video Dehazing And Electronic Image Stabilization Based On DaVinci Technology

The camera system such as handheld photography equipment, in-car camera system, aircraft orship photography platform had problems when working in the fog. On the one hand, the video wasnot clear because of the fog in the environment. On the other hand, the video was not steady for thereason of carrier shaking. It would lead to the observers’ misjudgment because of visual fatigue.Therefore, how to remove the blur and jitter was very important and urgent when the photographydevice working in the weather of fog.The paper researched the algorithms of rapid dehazing and image stabilization. It also focused onthe algorithm optimization and realization of the embedded platform based on the DaVinci technology.The algorithm treating in the YCbCr color space improved the speed of video processing embeddedsystem contrary to the video capture characteristics. In the part of video dehazing, we used shearhistogram equalization algorithm based on the enhanced algorithm to deal with the luminance channel.At the same time, we processed chrominance channels through a simple linear transformation. Inelectronic image stabilization part, we got the coarse matching using gray projection algorithm first,then we got precise matching using block matching algorithm which the block centers were thefeature points in the video. During precise matching, we introduced random sampling consensusalgorithm to remove mismatching points which majority were caused by the local motion in the video.By using random sampling consensus algorithm, we avoided the interference of local movement onthe image stabilization. We calculated the motion compensation parameters through the featurematching points using the method of least squares algorithm. After kalman filter, we got the finalmotion compensation parameters which would compensate the video using bilinear interpolationalgorithm. The compensated video without jagged lines and ripple was the stable video sequence.Finally, we transplanted the algorithm to the development platform of SEED-VPM6467T. Thepaper made full use of the hardware advantages for DM6467T. We used direct memory accesstechnology, caching technology and code optimization to improve the real-time of the system. Thespeed for processing720P video had up to20frames per second. This speed met the real timerequirement for video processing.