Multiple Images Fusing and Tone Mapping Algorithm

• Main Authors: Hsin-kai Ma, 馬欣愷
• Other Authors: Mei-chun Lo
• Format: Others
• Language: zh-TW
• Published: 2011
• Online Access: http://ndltd.ncl.edu.tw/handle/69041306406117555168

碩士 === 世新大學 === 資訊管理學研究所(含碩專班) === 99 === A multi-image fusion model, similar to the method suggested by Kao et al (2008) was firstly applied in this research to create high-dynamic-range (HDR) images. It was applied in the fusion process of three differently exposed images, captured by a typical still digital camera; and then in the creation process of HDR images, for further applications in the cross-media color reproduction. The images, created by the derived fusion method, had higher HDR than those of images obtained by conventional imaging process. Moreover, they also preserves much more details and information on both shadow and highlight areas, if compared to those images taken by the same digital cameras but using the conventional exposure method (i.e. only using one normal exposure). Additionally, since the dynamic range of most displays is limited, it would be impossible to obtain satisfactory representations of the original HDR scenes using such conventional soft-proofing displays. Therefore, a serial of algorithms were optimally derived, integrated, and tested in this thesis. These derived algorithms included local white-balancing for considering multi-illuminant scene conditions, tone-mapping, gamut-mapping, and CIECAM02 color appearance. Furthermore, a Gaussian pyramid method, which was based on a multi-scale model of adaptation and spatial vision, was also derived here. It was used to further enhance detail-rendition performance of the combined HDR imaging mechanism, derived in this research. Finally, the experimental testing results, (obtained from the imaging process of these integrated algorithms mentioned above), showed that the cross-media resulted images, when shown on the general display devices having lower HDR than those fused HDR images of interest, can pleasingly give satisfactory color appearances.