High Dynamic Range Image Synthesis with Image Registration and Deghosting

• Main Authors: Bo-JieTsai, 蔡伯杰
• Other Authors: Yen-Tai Lai
• Format: Others
• Language: en_US
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/86145661211725656524

碩士 === 國立成功大學 === 電機工程學系 === 104 === With the rapid development of modern technology, people pursue more realistic visu-al experiences. High dynamic range imaging, which can capture the approximate bright-ness range of the human eye, has been concerned. Because of the popularity of mobile de-vices, it is hoped that using mobile devices can also take high-definition pictures. Due to the expense of high dynamic range (HDR) sensors, many works try to generate HDR im-ages using several low dynamic range (LDR) images with different exposure conditions. Traditional HDR imaging requires that source LDR images must be static and without object movements. Mobile devices, often without a tripod, can’t prevent hand-arm tremors from happening. We developed a method which simulates the human visual system, mark-ing motion areas in the scene after alignment. Avoid any reference to these areas during recover radiance maps and keep the advantage of using multiple exposures. Experimental results show that the proposed method can achieve ghost-free HDR images for dynamic scenes. Compared to other methods, our method preserves more de-tails in motion area. It will help the development of HDR imaging in mobile devices in the future.