Abstract

Inverse Tone Mapping (ITM) methods attempt to reconstruct High Dynamic Range (HDR) information from Low Dynamic Range (LDR) image content. The dynamic range of well-exposed areas must be expanded and any missing information due to over/under-exposure must be recovered (hallucinated). The majority of methods focus on the former and are relatively successful, while most attempts on the latter are not of sufficient quality, even ones based on Convolutional Neural Networks (CNNs). A major factor for the reduced inpainting quality in some works is the choice of loss function. Work based on Generative Adversarial Networks (GANs) shows promising results for image synthesis and LDR inpainting, suggesting that GAN losses can improve inverse tone mapping results. This work presents a GAN-based method that hallucinates missing information from badly exposed areas in LDR images and compares its efficacy with alternative variations. The proposed method is quantitatively competitive with state-of-the-art inverse tone mapping methods, providing good dynamic range expansion for well-exposed areas and plausible hallucinations for saturated and under-exposed areas. A density-based normalisation method, targeted for HDR content, is also proposed, as well as an HDR data augmentation method targeted for HDR hallucination.

Highlights

  • High Dynamic Range (HDR) imaging [1] permits the manipulation of content with a high dynamic range of luminance, unlike traditional imaging typically called standard or Low Dynamic Range (LDR)

  • Where the predicted HDR image is denoted as ĨHDR and the Inverse Tone Mapping Operators (ITMOs) as f Inverse Tone Mapping (ITM)

  • The main procedure followed by ITMOs that are non-learning based is composed of the following steps [1]: 1

Read more

Summary

Introduction

The problem of inverse tone mapping is essentially one of information recovery, with arguably its hardest part being the inpainting/hallucination of over-exposed and under-exposed regions. In these regions, there is not sufficient information in the surrounding pixels of the LDR input for interpolation, compared to regions that are, for example, LDR only due to quantisation and preserve some colour and structure information. Inverse Tone Mapping Operators (ITMOs), attempt to generate HDR from LDR content. They can generally be expressed as: ĨHDR = f ITM ( ILDR ), where f ITM : [0, 255] → R+ (1). The operation can be local or global in luminance space

Methods
Results
Conclusion
Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call