Research On Large Field Video Mosaic Under Low-illumination Level

Language and image are the most efficient ways for people to obtain or exchange the information.The image information has been widely applied in both our daily life and the scientific study,and higher resolution and larger vision scope are the future trend of image.In the video security field,the traditional multi-camera surveillance system has a small vision scope by a single scene,and a bad continuity between different scenes.Meanwhile the surveillance images can be easily influenced by surrounding environment.For example,the quality of image may become poor at night as well as in somewhere light is not enough which may not meet the practical needs.Moreover,the picture with high quality and big vision scope can bring more useful information than the traditional picture.With the help of digital image enhancement and mosaic technique,people can get the high resolution and large vision video surveillance of the badly current scene,which has important theoretical significance and practical value.By analyzing the current low-lighting image enhancement method,this thesis aims to achieve a low-lighting image enhancement method by using the similarity between haze image and inverted low illumination image.This method is based on modified boundary constraint and image intensity.The algorithm works by first estimating the transmission function on the pseudo-fog image and then using the value component to get the automatic light.Extensive experimental results indicate that the proposed image enhancement method performs better than contemporary algorithms in terms of several metrics,including intensity,contrast with low complexity.Considering the current image mosaic algorithm handle image with large angular difference,the final stitching image may appear stretch distortion.This thesis proposed an image mosaic method based on best stitching plane.It reduces the stretch distortion by decreasing the transformation angular;traditional methods using weighting fusion can't remove the illumination difference,and cause seam-line and ghost problem.Byanalyzing HVS(Haman Vision System),an image seam-line removal method based on JND model has been proposed.Extensive experimental results indicate that the method can remove the seam-line efficiently.For low-lighting video stitching,this thesis combines low-illumination enhancement with video mosaic method and proposed the multi-scenario video mosaic method.The last low-illumination video test proves this algorithm can effectively enhance and stitch the video under the low-lighting or haze condition.