저 명암비 디스플레이를 위한 효율적인 대비 향상과 톤 매핑 기법

Abstract

The ultimate purpose of the digital imaging technologies is to reproduce an image which provides the perceptual match between the real world scene and the displayed image by digital devices. However, most digital cameras and display devices utilize the images with a smaller dynamic range than that of the real world scene. On the other hand, the digital imaging sometimes cannot cover the total available dynamic range. This is related to certain physical limitations of the image sensors and display devices. In spite of the improved imaging sensors and display devices, they are so expensive and unsuitable for small electronic image devices, such as mobile image applications. Therefore, the image processing including tone mapping and contrast enhancement are necessary for an appropriate dynamic range imaging according to the display devices. High dynamic range (HDR) imaging provides us a greater dynamic range of luminance to yield a more realistic scene. Since the HDR imaging contains a greater dynamic range than the displayed image, the dynamic range is compressed in order to reproduce the image on the display devices. Although HDR monitors are developed and researched, they are not generally commercialized and difficult to calibrate. Due to the reasons, a method called tone reproduction or tone mapping is necessary when using conventional display devices. In tone mapping, the HDR image contains the details in the extreme dim and bright regions. Thus, it is difficult to preserve the contrast of the details while compressing the HDR into the low dynamic range (LDR). Therefore, proper tone mapping techniques not only transform HDR images into the displayable range, but they also preserve the details, considering the features of the original HDR images. On the other hand, there are various reasons for poor contrast in the image and video. For example, a normal image has proper illumination/exposure, but some regions in the image are slightly under-exposed. Also, the backlight image is caused by over-exposed and/or under-exposed regions. A low-contrast image is caused by the insufficient illumination/exposure and a dark scene image in the night without a photoflash. These effects result in under-utilization of the offered dynamic range. As a result, the acquired images and videos may not reveal all the details and have a washed-out and unnatural look. Therefore, a proper contrast enhancement technique is necessary to obtain a more visually-pleasing image. This dissertation proposes two contrast enhancement algorithms and one tone mapping algorithm without causing the drawbacks in each method. The proposed technique basically focuses on enhancing the visual fidelity of both contrast enhanced image and tone reproduced image through the singular value decomposition (SVD). The contrast enhancement method for image aims to enhance the contrast while reducing the noise. The proposed method could overcome a common drawback of the previous contrast enhancements, which cannot successfully preserve the local edge contrasts. The experimental results demonstrate that the proposed algorithm outperforms the conventional methods. The contrast adjustment for stereo image is exploited to enhance the contrast while reducing the noise. The goal of the proposed technique is to obtain the identical luminance level from luminance inconsistency of the stereo images. The experimental results prove that the proposed algorithm achieves suitable visual quality for stereo images. Finally, tone mapping for HDR imaging proposes to enhance the contrast while enhancing the local details. Also, chromatic adaptation method is applied to preserve the chromatic appearance model. In the experimental results, the proposed technique preserves the local edge contrasts better while reducing the dynamic range appropriately. The experimental results of three proposed algorithms outperform the previous methods. Therefore, the proposed methods provide proper contrast image and preserving the details at the same time.