Research Of Generated Methods For High Dynamic Range Video

The dynamic range perceived by the conventional digital camera falls behind by far that of the human eyes.Therefore,human can't obtain the real scene information exactly due to the loss of detail and color distortion in the captured images or videos.The High Dynamic Range(HDR)images or videos not only show the exquisite color information,but also show the clear brightness level,which are consistent with the Human Visual System(HVS).Thus,the research of generated methods for high dynamic range video are studied in this paper.At present,There are mainly two high dynamic range video generation methods.One method is considered from the hardware point of view,HDR video is generated directly by the camera with a specific sensor,but the hardware of the method is expensive and not widely used.The other is considered from the software point of view,the most of HDR videos generated methods are based on alternating exposures for every LDR frame.Although this methods can produce a better quality and clearer HDR video,the input device of the methods has certain requirements.In order to meet all the equipments,this paper presents the generated methods for HDR video based on LDR video of normal exposure.First,the first frame image in the HDR video is generated by the generated methods for HDR image with a single LDR image.Then,this paper uses the correlation between adjacent frames to generate the HDR video.In this paper,two generated methods for HDR video are proposed which are based on motion estimation and curve fitting,respectively.In this paper,the contrast information and other factors are added when seeking weight maps in the generated methods for HDR image with a single LDR image.Besides,the pyramid decomposition method is used in this algorithm which can make the generated image clearer.The proposed method played a certain effect on HDR video generation.In the generated method for HDR video based on motion estimation,the first frame image in the HDR video is generated by the proposed generated methods for HDR image with a single LDR image at first.Then the motion estimation algorithm is used to find the moving region binary map and the static region binary map between the adjacent frames.Then the dynamic HDR regions and static HDR regions are generated.In the end,the HDR video is generated.In the generated method for HDR video based on curve fitting,the first frame image in the HDR video is generated by the proposed generated methods for HDR image with a single LDR image at first.Then the optimal fitting curve is builded between the the first frame image in the LDR video and the first frame image in the HDR video.The optimal fitting curve is used to generate the HDR video in the end.In this paper,the existing high dynamic range image quality evaluation methods are used to evaluate the proposed methods.The experimental results show that the proposed methods have better subjective and objective effects.