GPU-Based Real-Time 4K RAW Workflows

Advances in digital imaging technology are fundamentally changing the cinema workflow and the tools artists and engineers traditionally use. Relatively inexpensive 4K resolution digital motion picture cameras capable of capturing and storing RAW sensor data with a wide dynamic range, high color gamut, and high bit depths all at frame rates that have traditionally been the domain of broadcast video are now available. Implementing a RAW workflow that provides real-time interactivity and a production path where all artistic choices are non-destructive requires a great deal of compute as every image displayed needs conversion from RAW sensor data to display oriented imagery and colorimetry. This highly parallel operation is well suited to the capabilities of modern graphics processing units (GPUs). This paper will present best practices for optimal GPU compute core and memory usage as well as efficient data transfer schemes for sensor data processing and display.

Published

2012-10

Content type

Original Research

DOI

10.5594/M001433

ISBN

978-1-61482-952-2