What I want to do
I have converted a Canon Powershot ELPH 300 HS to use a wratten25A filter and I'm trying out, taking some photographs from vegetation places here and there. I'm evaluating how well these results fit my expectations and I notice some interesting results.
My attempt and results
Here's two photos of a place in Brazil. It has a pool, stone tile area, wooden crossings, vegetation, plants, flowers and some shadows. I calibrated the white balance on a red surface in 4/8 cloud cover, not in bright sunlight.
Questions and next steps
I especially notice a very high correlation of high NDVI values in shadowy areas, which could indicate that the radiation of NIR vs. VIS light is not linear. Because of the formula of (NIR-VIS)/(NIR+VIS) and assuming that NIR == 1, then it's clear that when VIS approximates 0 (towards shadowy areas), you'd expect a higher NDVI. In the results of the images above you can clearly see how strong this correlation works out in practice.
The next question is then why this occurs.
- Is my white balance still off? Should I find a better surface to calibrate against which is also directly sunlit?
- Does NIR bounce around more than VIS light?
- Is the sensor oversensitive to NIR?
Another interesting observation is that even at night I'm getting consistent 'blues' on the sensor, especially when I aim the camera at wooden parts like a wooden dresser over here, a wooden railing and so on. At night I wouldn't expect a lot of NIR light to enter the camera, so I'm wondering if I need to do something special to get better results.
It also seems to have a strong correlation with green items that come out very blue, like a green towel hanging around here.
Why I'm interested
Obviously modified cameras need to demonstrate results where there is a strong correlation between reflected VIS/NIR light in such a way that it doesn't have such strong correlations with the shadow. So I need to find out what's still wrong in the setup. Could still be the white balance, could also be some other settings in the camera. Any help is appreciated.