Recovery And Reconstruction Of Target Shape In Overexposed Video Stream

Image processing has a wide range of applications in many areas of social life.Image enhancement is an important sub-area in image processing,which refers to computer processing to improve the visual effect of images to meet the requirements of different occasions.In terms of image enhancement,domestic and foreign scholars have proposed a number of algorithms to enhance the visual effects of images to meet the different needs of people.The tracking and recognition of moving objects in visual images is also a subject of extensive research in image processing.For routine air and sea monitoring or airspace monitoring,it is often necessary to "peg" the moving target,and to accurately position the target,and sometimes to analyze and understand the shape of the target.In the continuous monitoring process of moving targets in an open environment,with the different orientations of moving targets,over-exposure often occurs in the video stream,which is not conducive to tracking and observing targets.In daily life,when the driving recorder or the surveillance camera is exposed to strong light,it will also be distorted or lose part of the image information.Now the research on over-exposure area detection and repair is on a single-frame picture,and no over-exposure video stream is applied.in.Therefore,there is a real need for the recovery and reconstruction of the target shape in the overexposed video stream,which is a relatively new research topic.This paper mainly studies the rapid recovery and reconstruction of target morphology in overexposed video streams.The main research content is divided into four parts:The first of all is tracking of targets in overexposed video.For the tracking of targets in overexposed video,the over-exposure light is full of complex and complex tracking backgrounds.The most classic KCF tracking algorithm in the discriminant method is used.The experimental results show that the KCF algorithm tracks the targets in the overexposed video stream.Has a good tracking effect.At the second is detection of overexposed areas of overexposed images.In order to detect the overexposed area of the overexposed image more accurately,this paper proposes an L2 regularized logical nonlinear regression(FLA)algorithm which combines various features as an improved algorithm,introduces the color features of the image area,and corrects the saturation.Feature,bright feature and boundary neighborhood feature,and L2 regularized logical nonlinear regression algorithm is used to obtain the classifier model through training to realize the detection of image overexposure region.Compared with the brightness threshold method and the LC algorithm,the algorithm can make the detection result of the image area more compact,and obtain the over-exposed area with good connectivity.The single over-exposure point will be less,and it is more consistent with the human eye to the exposure area.The third is over-exposure image correction and texture restoration.This paper proposes a repair algorithm for overexposed images based on improved texture synthesis.The original color image is first converted to the CIELab color space,and the L channel,the a channel,and the b channel of the image are separated.the L-channel of the image is decomposed into the structure layer and the detail layer by using the bilateral filtering algorithm.Then,the improved texture-based texture layer is used to synthesize the texture detail layer of the overexposed area.The effect of the method is verified by simulation comparison experiments.The results show that the method is feasible,it does automatically and effectively correct the overexposed image and produces vivid details in the overexposed areas.The last is recovery and reconstruction of overexposed video streams.For video reconstruction,the overexposed area detection algorithm based on the modified saturation feature is firstly used to detect the overexposed area of one frame of image,and then the overexposed image repair algorithm based on the improved texture synthesis is used to repair the overexposed area of one frame.The process is applied to each frame of the overexposed video stream to recover the reconstructed target shape of the overexposed video stream.