Natural Color Reproduction With Hybrid Tone Mapping And Perceptual Multi-exposure Fusion

The development of technology and the improvement of people‘s lives have led toan increasing demand for more accurate and pleasing visual experience; therefore, thenatural rendering of digital image has become a hot topic in recent years. High dynamicrange image processing is an advanced technique which could overcome the limits setby traditional display devices and gives our images a more natural look. Due to theabundant merits of high dynamic range image, it could be widely applied to computergraphics, medical images, remote sensing, surveillance, and so forth.This thesis focused on the major steps in high dynamic range imaging, namely,multi-exposure image fusion and tone mapping. Moreover, we improve these steps byintroducing perceptual qualities of digital image.In this thesis, we first proposed a novel high dynamic range image renderingtechnique with hybrid tone mapping and automatic k factor decision to automatically setall the parameters to avoid the parameter setting problem:(1) We combine global and local tone mapping operators, and manipulate the HDRcompression in both luminance and gradient domains to improve the appearance andnaturalness of results.(2) We also deal with the parameter setting problem effectively in HDRcompression by automatic k factor decision.We also proposed a perceptual multi-exposure fusion with overall quality index toimprove naturalness:(1) We propose the perceptual multi-exposure image fusion with overall imagequality index to choose the best image attribute to fuse in order to enhance the imagequality.(2) We also propose to apply color correction to the multi-exposure fused image tocorrect the color of the output images with the input images and the overall imagequality index was used as a measurement to select the reference image for colorcorrection.Experiments show that our methods can successfully enhance image quality andresult in a more natural look, therefore can be used in a variety of occasions in digitalimaging and video processing techniques. This work is supported by the National Natural Science Foundation of China (No.61271298）.