Shadow Mitigation

Shadows are present in a wide range of aerial images from forested scenes to urban environments.  The presence of shadows degrades the performance of computer vision algorithms in a diverse set of applications, for example.  Therefore, detection and mitigation of shadows is of paramount importance and can significantly improve the performance of computer vision algorithms.  This work assumes as input a multispectral image and co-registered cloud shadow map which are used to calculate shadowed pixel spectral statistics and adjusting to match the statistics of spectrally similar sunlit pixels resulting in a shadow mitigated multispectral image.

RIT is developing a shadow mitigation algorithm which focuses on removing the appearance of shadows in high spatial resolution multispectral overhead collected imagery. The intent is for material spectral appearance to be constant regardless of the level of sun-light and shadow for use in sponsor algorithms. Our approach utilizes multispectral image data and a binary cloud shadow map as input and outputs a multispectral image with shadows removed. The removal of shadows is accomplished by determining spectral statistics of shadowed materials, comparing to spectral statistics of spectrally similar sunlit material surfaces and adjusting the shadowed pixels to match the brightness of the sunlit pixels. We have found good shadow mitigation results for small (512x512 pixels) image chips and less than optimal results for larger scene extents. We are currently researching methods for processing statistics on a cloud by cloud basis to extend the acceptable image chip performance to an acceptable full scene performance.