High Dynamic Range Image Generated Based On Multi-Exposure Images

As the real scene contains a large dynamic range and has a broad range of colors and intensity information. To some extent, we can obtain details of the real scene by our human visual system, but not recorded by some traditional display devices because of the limitation of CCD and ADC. The display devices only can exhibit some local scene information, such as contrast, intensity and colors, which is not comparable with human visual system. When the dynamic range of a scene is wider than a normal camera, photographers will often change camera exposure in order to control the captured range of brightness, which results in the loss of some details as some under-exposed or over-exposed regions in a picture. Some methods are proposed in this paper for fusing multi-exposure images of a static scene are taken by a stationary camera into a high quality image. The concrete contents are shown as follows:(1) The multi-exposure images are first divided into some blocks, and then three different measure factors are selected to obtain the highest information block. A fusion function is introduced to eliminate boundary discontinuities. The resulting image can show more details than one of source images.(2) The multi-exposure images are decomposed by pyramid-based method, the sub bands are combined with some weight maps, guided by three measure factors. The final fusion image can be obtained by pyramid reconstruction.(3) A method for decomposition of multi-exposure images by discrete wavelet transform or discrete wavelet frame is proposed. The low frequency sub bands are fused based on weight maps. The high frequency sub bands are combined with local energy information. The inverse discrete wavelet transform or frame is applied to obtain the final image.(4) To assess the color fusion image quality, an approach is presented for color image fusion using a symmetric similarity function.The methods presented in this paper are effective to obtain high dynamic range image. The resulting image contains all the details in the original images, which can improve the display result for traditional devices and is more adaptive to the future research in computer vision.