自由无拘束多曝光融合的研究

自然界中光线的亮度跨越十多个数量级，远远超出了当前成像设备的动态范围。所以我们拍摄的影像经常出现过爆或欠曝问题，因而被称为低动态范围影像。为了更准确的呈现场景中信息，高动态范围影像势必为未来趋势。基于多曝光融合的高动态范围技术近年来引起了学界的广泛关注，并成功的应用在一些新产品中如iPhone4。然而，传统算法计算量大，而且对目标场景有较高的要求，更无法在运动、模糊、低照明等场景中获取有效的高动态范围影像。本项目拟针对当前多曝光融合算法的缺陷，在保证低计算量的基础上提高其处理复杂环境的能力。我们将避开传统算法所需的相机光度标定、色调映射等步骤，完全从图像融合的角度探索噪声、运动的影响。算法研究重点是多曝光图像的配准、去噪、去模糊、去鬼影等问题。由于针对的图像多存在过曝或欠曝问题，所以较传统的图像问题更具挑战性。改研究将全面提高多曝光融合理论及算法的实用性，使用户可以自由无拘束的操作。

Radiance of the real world spans several orders of magnitude and its dynamic range dramatically exceeds the capability of the current electronic imaging devices. As a result, the images we captured can be referred as low dynamic range (LDR) images and offen suffer from over- or under-exposure problems. Apparently, high dynamic range (HDR) image is the future trend as it can encode the world with higher color precision. The multi-exposure techniques have aroused much attention recently, and soon the feature became available in some commercial imaging products like iPhone4. However, conventional algorithms are of high computational cost, and cannot produce desirable results when the target scene is dyanmic，blurring or noisy due to low lighting condition. The target of this project is to overcome this limitation and make exposure fusion applicable in complex scenes while keeping low computational cost. This project will address the influence of motion and noise to exposure fusion in image domain. Thus, typical HDR steps such as photometric calibration and tone mapping can be ignored. The focus is to investigate the registration, denoising, deblurring and deghosting for multi-exposure images. Apparently, these tasks are more challenging than those in general images, since the images here are differently exposed and may have over- or under-exposure problems. This work will improve the practicality of exposure fusion greatly, and let users enjoy the fun of HDR more easily.