Color balancing algorithms such as histogram matching are often applied to remote sensing data to make imagery of the same area, captured at different times, appear visually consistent with a reference image. However, when the input data differs from the reference image because of clouds, snow, or other issues, existing color balancing methods can produce severe artifacts. Planet introduces a new method, implemented using the SciPy stack, that fits a smooth color transfer function based on coregistered points and prior knowledge of the approximate white balance for the image. This approach provides color consistency with the reference image without introducing visually unrealistic artifacts when clouds and snow are present.
Watch the robust color matching talk on YouTube.
Comments
Article is closed for comments.