Research On HDR Video Acquisition Technology

High Dynamic Range(HDR)video acquisition technology mainly refers to the HDR video generation technology of extending Low Dynamic Range(LDR)video to HDR video.It has been applied in many fields such as photography,film factory,game and security and so on.The existing HDR video generation methods can be divided into two categories: one is based on single-frame images,the other is based on multi-frame images.We mainly study on the single-frame based methods of HDR video generation in this thesis.The main problem which existing single-frame based methods of HDR video generation focus on is expanding dynamic range through making use of brightness information of current frame.Some methods do not take into account time consistency of brightness between video frames.Some methods do not take into account computation complexity.Therefore,in this thesis we focus on three points: first,the HDR image generation method which can appropriately expand the dynamic range of the image and keep the contrast details of original image;Second,HDR video generation methods which can maintain the time consistency of video frame.Third,the method of real-time HDR video generation.At first,we propose a HDR image generation method based on MSR(Multi-Scale Retinex).We decompose an image into two images based on MSR method,and propose two different enhancement function according to the characteristics of these images: the guidance filter enhancement function and the gamma map enhancement function.Secondly,we improve the proposed HDR image generation method by adding scene change detection,flicker detection and deflicker methods into the original methods,and an HDR video generation method that can maintain the time consistency of brightness between video frames is achieved.We have studied and improved different scene change detection methods,and proposed a scene change detection method that guarantee the detection recall and precision ratio.We also have studied flicker detection methods in LDR videos.Through a large number of experiments,we have determined suitable flicker detection methods and related detection parameters for HDR video.Finally,we improve the HDR video generation method proposed in this thesis by replacing the previous HDR image generation method with a new guided filter based HDR image generation method in order to reduce the computation complexity,and make it applicable to real-time applications.We have determined the best performance of enhancement function based on guided filter by experiments.In addition,in this thesis we evaluates and compared these proposed methods with other HDR image generation methods and video generation methods from subjective and objective aspects.Experimental results show that the MSR based HDR video generation method proposed in this thesis can obtain high quality HDR video frames and maintain the time consistency of brightness between video frames,the real-time HDR video generation method proposed in this paper has a certain compromise between the image quality and computation complexity,and it can be applied to the real-time HDR video generation.