基于动态场景的高动态图像合成

相机所取得的图像动态范围与真实的场景相比往往较低。而通过对几幅不同的曝光时间的图像进行合成,可以得到高动态图像。但是场景中物体的运动会导致最终图像出现鬼影效应,提出了一个基于区域分割和连续性检测的算法,消除鬼影区域,使得在场景有运动下同样可以获得高质量的高动态图像。

High Dynamic Range Image Exposure Fusion in Motion Scenes

The dynamic range of Images captured by normal camera are often not high enough comparing with real-world scenes. A high-dynamic-image can be acquired by fusing several images with different exposures. But direct fusion may cause ghost effect when motion exists in the scene. This paper proposes a framework to eliminate ghost-effect based on region segmentation and consistency detection, making it possible to achieve ghost-free high-dynamic-image even in motion scenes.