High Performance Image Processing Algorithms for Current and Future Mastcam Imagers

In this project, Applied Research LLC (ARLLC), the University of Tennessee, Knoxville (UTK), and the Arizona State University (ASU) propose high performance image processing algorithms that will support current and future Mastcam imagers. The algorithms fuse the acquired Mastcam stereo images at different wavelengths to generate multispectral image cubes, which can then be used for high quality virtual reality (immersive) visualization in 3D, data clustering, anomaly detection, and rough composition estimation from relatively long distance when compared to LIBS instrument. One major challenge in constructing a multispectral image cube from the two Mastcams is the alignment of the images in the stereo image pair which needs to have registration errors in the subpixel level. To address the challenge in the stereo image alignment, we propose a two-step image registration framework. In this framework, we also provide a set of image processing techniques, including pansharpening, debayering, data clustering, and anomaly detection. In Phase II, we will further validate the above algorithms using Mastcam-Z data. Moreover, we will develop new data products for generating 12-band image cubes, high resolution stereo images, and new layers for Java Mission-planning and Analysis for Remote Sensing (JMARS).