GPU Focused Image Compression Techniques

Recent advances in graphics hardware have placed a dedicated graphics chip in every pocket. However, this processing power is limited by restrictions on energy and bandwidth. In order to work within these constraints, it is imperative to reduce the amount of data we transfer. For image-heavy applications, such as geographic information systems or simulated environments, the classic way to reduce the amount of data is to use image compression. Traditional compression approaches do not map well to a GPU architecture, and we must develop clever techniques that work within these limitations. This talk will cover the requirements of a good GPU image compression format and describe some of the current formats available. Additionally, the talk will cover algorithms developed to encode images into these formats and some of the future directions that may be promising.

Speaker Bio: Pavel Krajcevski is a fifth year graduate student at the University of North Carolina at Chapel Hill where he studies GPU compression techniques under Dinesh Manocha. His work has been applied at a number of companies from Samsung and Intel to Google. Prior to graduate school, Pavel spent two years at Disney Interactive Studios as a software engineer, and holds Bachelor's degrees in both Computer Science and Mathematics from the University of Chicago.

Presentation Slides: PPTX

Note: Visitors from outside NIST must contact Cathy Graham; (301) 975-3800; at least 24 hours in advance.