Taeuk Kang, Jeonghyun Lee and Jechang Jeong, Hanyang University, Korea

In order to display HDR (High Dynamic Range) images with increased dynamic range on LDR (Low Dynamic Range) monitors, it is necessary to perform a tone mapping technique, which is a process of compressing a dynamic range of an image. Representative techniques are global tone mapping and local tone mapping. Though global tone mapping is simple to compute, it has low local contrast and loses details. Local tone mapping has high local contrast but it demands high computational complexity. In order to take advantage and to compensate of two techniques, we propose local tone mapping based on clustering. Clustering reduces the complexity of local tone mapping implementation by dividing images. In the local tone mapping process, the local adaptation is obtained by combining the cluster-level log mean and global log mean. Using local characteristics, that is local adaptation, based on clustering, the results has high local contrasts and local detail is improved. Experiment result shows that proposed algorithm has better performance than conventional algorithm.

High Dynamic Range(HDR) Imaging, Tone Mapping, k-means Clustering