Modifying the NDVI Values Using Multispectral Images

Can I determine the minimum and maximum value of the NDVI variable? How do I make a threshold for example higher or equal to 0.3 are green and lower than that are red? Can I make bounding box for the specific pixels that has low values?(but in the plant area, not the road or things that are not vegetation)

Below is the code I used to compute :

R = imread('RED.TIF');
NIR = imread('NIR.TIF');
NDVI = (double(NIR) - double(R)) ./ (double(NIR) + double(R));
imshow(NDVI, []);

Answer 1

[Image Analyst]

Try this:

maxValue = max(NDVI(:))
minValue = min(NDVI(:))
binaryImage = NDVI > 0.3;

You might also be interested in my interactive thresholding app in my File Exchange.

[Thomas Taufan]

What if I want it to be 3 colors rather than binarized image that is black and white? For example NDVI valued -1 to 0 would be black, 0.1 to 0.39 would be red, and 0.4 to 1 would be green?

[Image Analyst]

Several ways to do that.

1. One is to create three masks and put them into the correct color channels of an RGB image.
2. The other is to just pseudocolor the image by using a colormap. Use colormap() and caxis().

The first method is probably easier

binaryImageR = NDVI > 0.1 & NDVI < 0.39;
binaryImageG = NDVI > 0.4 & NDVI <= 1;
black = zeros(size(binaryImageR), 'uint8');
rgbImage = cat(3, uint8(binaryImageR * 255), uint8(binaryImageG * 255), black);
imshow(rgbImage);

[Thomas Taufan]

How can I save the NDVI variable using imwrite?

If I used this imwrite(NDVI,'NDVI.jpg') the result saved is different from the plot seen in imshow(NDVI,[])

What is the [] used for? Thanks

[Image Analyst]

To save as a double in a .mat file:

save('ndvi.mat', 'NDVI');

To save as an image that other programs can use, convert to uint8 or uint16:

img8 = uint8(255 * NDVI);
imwrite(img8, fullFileName);

The [] is used to scale the actual range of the image -- 0 to 1 in your case -- to the full 0-to-255 range that your display adapter wants.

[Thomas Taufan]

Thanks for your kind attention and helpful answers, I would like to ask one last question please I used imread to read the pixel value of the images, can I read the value for example at specific wavelength 720nm or 680nm? The camera for example captures red at range 620nm to 750nm.

[Image Analyst]

Not if it's a regular RGB camera. You'd need a hyperspectral camera for that.

[Thomas Taufan]

Could you kindly explain to me why is that?

What does the values of the variable using imread mean?

I have this image and I want to split the channel using this command

Original_Picture = imread ('Picture 80.jpg');
R = Original_Picture(:,:,1);
NIR = Original_Picture(:,:,3);

And then I compute the NDVI value, but the value I got is very low, and after reading some references it is because of the contamination of NIR in the Red channel. How can I reduce that or in other way can I choose the specific wavelength of the R and NIR variable to be read?

[Thomas Taufan]

Hello Image Analyst, could you please take a look at my question? Thank you

[Image Analyst]

The values from imread() are assumed to be the RGB values of the image. Like you said the R channel goes from 620 to 750 which seems a bit more IR than normal. But it's possible if you have a camera where they changed or got rid of the IR cutoff filter that is normally on top of the sensor. But a normal DSLR or machine vision camera would not have sensitivity out to 750 nm, only to about 700 or so. From 700 to 750 is near IR - very near. So near that it's actually missing most of the band that people would call near IR which is roughly 700 or 750 to about 1200-1400 nm. So band 3 which you extracted from Original_Picture is actually mostly red, unless you have a special camera to filter out the red and allow only wavelengths greater than 700 in. To get better IR you need to get a camera that is specifically made to record IR wavelengths. There are cameras like this available. I think you might even be able to get someone to alter your Nikon camera to remove the IR filter from the sensor or replace it with a sensor that doesn't have the filter on it.