Spatially Adaptive Histogram Equalization for Single Image-Based Ghost-free High Dynamic Range Imaging

Abstract

In this paper, we present a spatially adaptive histogram equalization method for generating a ghost-free high dynamic range (HDR) image using a single input image. Existing multiple input-based HDR methods fuse multiple low dynamic range (LDR) images, acquired using different exposures. However, these methods work only under the assumption that neither global nor local motions exists in between the LDR images. To overcome such an unrealistic constraint, we generate two LDR images from a single input image. To generate LDR images with appropriate exposures, we divide the entire intensity range into multiple sub-ranges using histogram quantization and separately perform histogram equalization in each sub-range. Thus, we can acquire a set of differently exposed LDR images of the same scene, which are then fused to generate a ghost-free HDR image. The major contribution of this work is twofold: (i) a novel estimation method for providing optimal sub-ranges of intensity using histogram quantization, which (ii) requires no additional hardware for generating multiple LDR images. Because the proposed method uses a set of optimally self-generated LDR images, it is inherently free of ghost artifacts and can provide a ghost-free HDR function for low-cost, lightweight imaging devices, such as mobile phones and compact cameras.