Filter Gaussians by Covariance #10
Description
Description
The input images are optimized against 8 independent Gaussian curves. Some curves may be left without a bright spot in the case of a capture resulting in fewer than 8 lit LEDs. Therefore the remaining Gaussians will start to diverge yielding a high covariance parameter. If we can filter the Gaussians by their covariance we can dynamically count LED bright spots which is important for LED calibration.