I've modified a Panasonic DMC-LZ5 digital camera to take photos for use in creating an infragram. I've removed the filter, and installed a red filter in its place inside the camera (filter material is Rosco E-Colour #019: Fire). Also, did the custom white balancing using the instructions by setting on a bright red object in direct sun.
The photos appear slightly turquoise as suggested they should be (perhaps not as turquoise as they should be), however, when I generate the infragram using the public labs website, I get the image as shown above. (the flower in the photo is actually a reddish/magenta color, so maybe my subject should've been a white flower)
In the original photo (top), you can see a shadow just under the flower in the photo, however in the infragram it shows more yellow than the surrounding area. Doesn't this imply more photosynthesis than the surrounding area? And shouldn't a shaded area be experiencing less photosynthesis? I also see this in the other photos, where the sun exposed areas are green, and the shaded areas in the depths of the plant foliage tend to be yellow or even red as in the processed image!
I've tried messing with the white balance with different reds. Also the histogram shown seems to have good separation between the red and blue channels, so I'm not sure what I'm doing wrong.
Any help you can provide would be greatly appreciated. I've been scouring the website and pulling my hair out on this!
I've also tried a Raspberry Pi Noir camera with the same filter, and it seems to produce more appropriate images, so I'm wondering if I have some setting wrong, or there is just something inappropriate with the panasonic camera sensor I'm using.
It's a common artifact to get higher NDVI values from the darker areas of foliage in DIY infrared photos. The computation of NDVI depends not only on the difference between the NIR and red light reflected from foliage (the NIR-red part), but on the absolute value of the NIR and red values (the divide by NIR+red part). Darker areas of foliage (e.g., shaded) could have the same difference between NIR and red, but a much lower sum of NIR and red. That would result in higher NDVI values. For example, (200-100)/(200+100) computes to NDVI=0.33, but (150-50)/(150+50) computes to NDVI=0.50. There are probably some other similarly devious artifacts that cause unrealistic results when deriving NDVI from DIY infrared photos.
So sometimes the results can be better on cloudy bright days when there are no distinct shadows. It's also important to capture photos that are well exposed. Overexposure shifts the histograms to the right, and underexposure shifts them to the left. Either shift will probably result in different NDVI values. This is one reason that it is difficult to compare the NDVI results from two different photos -- they must be exposed exactly the same (light quality and angle must also be the same).
Above is an NDVI image made in Fiji. No adjustments were made. The NDVI values for foliage are around 1.5 to 2.5, and more typical values might be 4 to 6. A custom white balance which uses a color that floods the sensor with more reddish/purplish light could produce higher NDVI values.
Cameras might differ a lot in their relative sensitivity to red and NIR. So that critical relationship between the brightness of the red and NIR will vary depending on the camera used.
Chris
Reply to this comment...
Log in to comment