A unified postproduction workflow for low-light acquisition environments

Pui Anantrasirichai

Genres such as natural history place extreme demands on camera technologies and postproduction due to their challenging acquisition conditions. Distortions are particularly endemic in low-light video due to the interactions between lens aperture, shutter angle and ISO setting. This project explores perceptually inspired solutions to a class of inverse problems associated with video production: denoising, colorization and contrast enhancement, ensuring temporal consistency in terms of colour, luminance and motion. Training strategies and databases are needed that are cognisant of acquisition context, capable of learning without explicit ground truth information. This will need perceptually accurate objective quality metrics and loss functions to validate our methods in the context of professional content production and delivery workflows.