Fused Color Night Vision Imagery Using Invariant Spectral / Texture Identification within a Bayesian Network.

The proposal addresses the need for better multi-camera fusion through the development of novel material identification and image processing routines. An innovative spectral/texture algorithm adapted specifically for night vision imagery will be used to segment the scene. Different portions of the scene are then identified using a Bayesian Network that utilizes ancillary knowledge to "intelligently" manage an invariant database. Finally, color and lightness values are assigned to maximize scene understanding and object identification. The algorithmic framework is versatile for handling different combinations of sensors operating in different spectral ranges, and it is modular so that it can utilize additional algorithms where appropriate. The overall approach builds on previous OKSI work in true-color night vision, multispectral image processing, infrared scene rendering, and multi-camera fusion. It moves beyond the current state of the art by using innovative algorithms that enable efficient and invariant scene rendering with true-color displays. Under Phase-I, OKSI will develop all aspects of the project and demonstrate the algorithms on both real and simulated night vision imagery. Under Phase-II the database will be expanded and the algorithms will be optimized for performance and speed, and a fully functional prototype fusion system operating in real-time will be developed.