One-Day Outdoor Photometric Stereo Using Skylight Estimation

摘要

We present an outdoor photometric stereo method using images captured in a single day. We simulate a sky hemisphere for each image according to its GPS and timestamp and parameterize the obtained sky hemisphere into quadratic skylight and Gaussian sunlight distributions. Our previous work recovered an outdoor scene on a clear day, whereas the current paper shows that cloudy days can provide better illumination conditions for surface orientation recovery, and hence we propose a modified sky model to represent a well-conditioned skylight distribution for outdoor photometric stereo. The proposed method models natural illumination according to a sky model, providing sufficient constraints for shape reconstruction from 1-day images. We tested the proposed method to recover various sized objects and scenes from real-world outdoor daylight images and verified the method using synthetic and real data experiments.