If your objective is just to indentify vegetation class, then you can simply threshold the NDVI. Lets say all the pixel value greater than certain value is vegetation. For determination of the threshold you can simply use interactive trial and error approach or there are many histogram based automatic thresholding techniques.

[QUOTE]shreshai wrote If your objective is just to indentify vegetation class, then you can simply threshold the NDVI. Lets say all the pixel value greater than certain value is vegetation. For determination of the threshold you can simply use interactive trial and error approach or there are many histogram based automatic thresholding techniques. [/QUOTE] Thank you. Can I load the NDVI as one of the bands and then run a classification?

[QUOTE]tarsierspectral wrote shreshai wrote If your objective is just to indentify vegetation class, then you can simply threshold the NDVI. Lets say all the pixel value greater than certain value is vegetation. For determination of the threshold you can simply use interactive trial and error approach or there are many histogram based automatic thresholding techniques. Thank you. Can I load the NDVI as one of the bands and then run a classification? [/QUOTE] You dont have run any sort classification necessarily. You can just use band math in ENVI to get a binary vegetation map. If you are using any supervised classification, you have to collect samples and incorporation of NDVI might help you to get better representation of vegetation class. In that case, you simple append NDVI in the image data, and run the supervised classification with the samples you have collected. If you explain what you are trying to do, it will be easy to suggest.

If you want to run an unsupervised classification, you would probably want to use your original data, not the NDVI. Unsupervised classifications find spectral clusters in the data, and identify them as separate classes. NDVI is just one image, not a multi-spectral dataset, so it can't be used as input to an unsupervised classification. You could add the NDVI to your multi-spectral dataset, as an additional band. But then you will really be comparing apples to oranges in your classification. The NDVI band is measuring a different thing than the other bands of the data, and it is likely to have a different range of values, etc. which could end up skewing your classification in an unhelpful way. On the other hand, if you want to use the NDVI values only, as the basis of your classification, then you can do as suggested previously. Just threshold your NDVI image, so that low NDVI values are assigned to one class, medium to another, and high to another (or however you want to divide them up). One easy way to do this in ENVI 5 is by displaying the NDVI image, then right clicking on its layer in the Layer Manager, and choosing Raster Color Slices. This tool lets you apply different colors to pixels with different NDVI values. I hope that helps. As the previous poster mentioned, if you can give more info about what your ultimate goal is, and what data you are working with, you may get more specific advice. Peg Exelis VIS

Thank you for all the suggestions so far. What I am trying to do is show vegetation change over a period of time. I have 2 Landsat TM images, one from 1984 and one from 2007. I want to show how the vegetation changed from 1984 to 2007, whether it increased or decreased. I don't need to identify different vegetation types. How would I approach this?

You may want to calculate NDVI for both dates, then use ENVI's Change Detection Difference Map with the NDVI images as the Initial State and Final State input images, to find the areas that have more and less green vegetation. - Peg Exelis VIS

[QUOTE]Peg wrote You may want to calculate NDVI for both dates, then use ENVI's Change Detection Difference Map with the NDVI images as the Initial State and Final State input images, to find the areas that have more and less green vegetation. - Peg Exelis VIS [/QUOTE] Hi Peg, Thank you very much. Once I create the difference map is there some kind of explanation how to interpret it? I see 11 classes going from -5 to +5. 0 indicates no change, do the negative numbers indicate vegetation loss and positive increase? What about simple difference vs percentage difference? I really appreciate all the help you provided so far. Thank you

Yes, the explanations you are looking for are in the documentation for the tool, which you will find in the online help system, under Detection > Change Detection Analysis > Change Detection Difference Map. - Peg Exelis VIS

Thank you Peg. I am confused by the percentage parameter. I was hoping that it would give me a change by percentage. For example it would say that that areas in l dark red increased by 35% and let's say the ones on light blue decreased by 12%. What does "percent" mean in this case?

Percent is measured in terms of the pixel values here. So, if your pixel values are NDVI, it will be % change in NDVI value. - Peg Exelis VIS

[QUOTE]Peg wrote Percent is measured in terms of the pixel values here. So, if your pixel values are NDVI, it will be % change in NDVI value. - Peg Exelis VIS [/QUOTE] Thank you Peg