Precise Automatic Image Coregistration Tools to Enable Pixel-Level Change Detection, Phase II

Automated detection of land cover changes between multitemporal images (i.e., images captured at different times) has long been a goal of the remote sensing discipline. Most technology in this area has focused on methods for detecting and identifying land cover or surface object changes in two or more images, but precise co-registration of images remains a key challenge. In fact, image-to-image registration and image-based change detection are intricately related, as the success of conducting both relies on the precision of the other; software that supports these functions should do so in an integrative manner. Image registration is the key factor influencing the success of detecting land cover changes at or near pixel scale. We will develop tools in the form of a "software development kit" (SDK) specifically optimized for precise co-registration of two or more images with minimal user interaction, with the primary motivation to enable change detection algorithms to focus on salient changes rather than highlight image registration errors. The SDK will be available to NASA at no cost, after which we will build user applications based on the SDK for commercial offering.