Probabilistic exposure fusion.

Area of Science:

• Computer Vision
• Image Processing
• Computational Photography

Background:

• Natural scenes often exhibit high dynamic range (HDR), posing challenges for standard dynamic range (SDR) capture devices.
• Existing tone mapping operators struggle with preserving details in HDR scenes or introduce visual artifacts.

Purpose of the Study:

• To propose a novel scheme for capturing and synthesizing HDR scenes using SDR devices.
• To enhance scene detail and suppress gradient reversals in the synthesized SDR images.
• To provide a tone mapping solution superior to existing global and local methods.

Main Methods:

• Utilizing a series of SDR images with varying exposures to capture scene details.
• Calculating optimal image luminance levels and scene gradients from the SDR image series.
• Employing a probabilistic model to synthesize an SDR image preserving luminance and suppressing gradient reversals.

Main Results:

• The synthesized SDR image contains significantly more scene details than individual captured SDR images.
• The proposed scheme effectively functions as tone mapping, outperforming global and local operators.
• Subjective evaluations indicate user preference for the proposed method over existing approaches.

Conclusions:

• The novel scheme successfully captures HDR scenes with SDR devices, enhancing detail and visual quality.
• The method offers an automated and parameter-free solution for HDR to SDR image synthesis.
• This approach represents a significant advancement in tone mapping and HDR image processing.