Hi all,
I would like to know how to correct color and light in a single image when the deviations are not evenly distributed and they depend on the area of the image. I have hundreds of images taken with a GoPro camera from a plane. Their quality is not perfect but it is what we have. I found a strong effect on the color, and reflectance of the features in the image due to the perspective. So, because the sun light fall with an angle, the features on one side of the image looks over exposed while on the center and opposite side of the image the colors are more real, or at least natural, maybe a little darker towards the edge of the image. This effect is especially marked with vegetation, not only increasing the intensity but changing greens toward yellow.
I created the mosaic with Ortho Mapping and the color balancing tool wasn't able to correct those deviations. For me the problem is that same vegetation type appears with very different reflectance across the mosaic in function of what frame the Ortho-Mapping algorithm has decided to choose for this area. And worse, the overexposed areas can be similar although they correspond to different vegetation type. This effect had a strong influence in the classification and I need to remove. I guess this is something has to be corrected in each image individually before mosaic and balance them. However, the solution must be automated because the number of images to correct.
Thanks for all ideas and suggestions.
Solved! Go to Solution.
This effect is called Hotspot and occurs because the diffused reflection is maximum at the direction to the sunlight, so the areas of the image closer to the point where camera, sunlight and ground are aligned experience that over-exposure.
Apparently, the best method to reduce is to use a homomorphic filter.
This effect is called Hotspot and occurs because the diffused reflection is maximum at the direction to the sunlight, so the areas of the image closer to the point where camera, sunlight and ground are aligned experience that over-exposure.
Apparently, the best method to reduce is to use a homomorphic filter.