Deep Neural Network Algoritms for Upscaling of Surface Images
Physical Sciences Inc. (PSI) has been awarded a research program from the National Aeronautics and Space Administration (NASA) to develop the Single Image Super Resolution for Quantitative Analysis (QuantSISR) software suite comprising state-of-the-art super-resolution (SR) algorithms optimized to reduce errors during subsequent image analysis such as common computer vision tasks (image segmentation, object detection).
QuantSISR will be designed to achieve a 50% reduction in edge localization errors while matching pixel-wise accuracy comparable to methods optimized for visual perception quality. The software will be capable of 4x-8x up-sampling of satellite images based on image statistics learned during training and augmented with textures extracted from high resolution reference images. This feature can be used during Solar System exploration missions to mitigate mismatch between terrestrial training data sets and the newly acquired data. By leveraging multiple observation geometries, high resolution in situ references can be obtained and used to enhance wide area images acquired at lower spatial resolution.
PSI’s QuantSISR capability will directly address NASA’s need for more accurate super-resolution of existing and future observations. Potential NASA applications include Moon to Mars (rover navigation, obstacle avoidance); Europa Lander (landing site selection); long-term Earth observations (combining historical low resolution images with currently available high resolution images), such as Surface Biology and Geology (SBG) mission. Non-NASA Commercial Applications include up-sampling of low-resolution images to improve accuracy and/or reduce cost of analyses used for Land Management, Urban Planning, Environmental Monitoring, Transportation and other applications.
For more information, contact:
Dr. Bogdan Cosofret
Vice President, Detection Systems
Physical Sciences Inc.
Telephone: (978) 689-0003