Examining Autoexposure for Challenging Scenes

York University
ICCV 2023
*Equal Contribution

Abstract

Autoexposure (AE) is a critical step applied by camera systems to ensure properly exposed images. While current AE algorithms are effective in well-lit environments with unchanging illumination, these algorithms still struggle in environments with bright light sources or scenes with abrupt changes in lighting. A significant hurdle in developing new AE algorithms for challenging environments, especially those with time-varying lighting, is the lack of suitable image datasets. To address this issue, we have captured a new 4D exposure dataset that provides a complete solution space (i.e., all possible exposures) over a temporal sequence with moving objects, bright lights, and varying lighting. In addition, we have designed a software platform to allow AE algorithms to be used in a plug-and-play manner with the dataset. Our dataset and associate platform enable repeatable evaluation of different AE algorithms and provide a much-needed starting point to develop better AE methods. We examine several existing AE strategies using our dataset and show that users prefer a simple saliency method for challenging lighting conditions.



Download the 4D AE Dataset

Below are the download link of the 4D AE dataset. By clicking on 'Download' button, you will be directed to another link where you can download the dataset of all the 9 scenes. It contains 3 folders named as 'dng', 'sRGB_npy', and 'RAW_npy'. In the 'dng' folder, 9 sub-folders are listed where each of them has 1500 RAW images (6720 * 4480 pixels) stored in the dng formate. They are ordered as 100 frame * 15 exposure (15 s, 8 s, 6 s, 4 s, 2 s, 1 s, 1/2 s, 1/4 s, 1/8 s, 1/15 s, 1/30 s, 1/60 s, 1/125 s, 1/250 s, 1/500 s). The 'sRGB_npy' folder contains 9 npy files in the size of 100 frame * 40 exposure * 640 * 960 * 3, and 'RAW_npy' has 9 files for each scene in the size of 100 frame * 40 exposure * 1120 * 1680. The 40 exposure time includes 15 s, 13 s, 10 s, 8 s, 6 s, 5 s, 4 s, 3.2 s, 2.5 s, 2 s, 1.6 s, 1.3 s, 1 s, 0.8 s, 0.6 s, 0.5 s, 0.4 s, 0.3 s, 1/4 s, 1/5 s, 1/6 s, 1/8 s, 1/10 s, 1/13 s, 1/15 s, 1/20 s, 1/25 s, 1/30 s, 1/40 s, 1/50 s, 1/60 s, 1/80 s, 1/100 s, 1/125 s, 1/160 s, 1/200 s, 1/250 s, 1/320 s, 1/400 s, 1/500 s.


>

Results Visualization

The following examples show a comparison of 4 AE methods (Global, Semantic, Saliency, and Entropy) evaluated on the 4D AE dataset.

Scene Selection

AE methods Comparison

Saliency (Ours)
Global
Semantic
Entropy

Frame Control
Play
Pause


Sample Data

The following examples show 2 sample datasets with 15 exposure values in 20 time steps..

Scene Selection

Dataset Viewer


Shutter Speed