Is Histogram Equlization is necessary in This case?

Based on my experience you should not use image enhancement techniques when computing NDVI. However, if you compute NDVI of time series images, atmospheric correction may be required.

You should not use image enhancement at all to compute NDVI.

Convert the DN into Top of air reflectance before calculating NDVI.

How to covert the DN into Top of air reflectance ..please mention

How to covert the DN into Top of air reflectance ..please mention

You can use Radiometric calibration tool available in ENVI, or read User Guide of the image used to find the conversion equations, then you can write your own code to do that.

Example: http://www.yale.edu/ceo/Documentation/Landsat_DN_to_Reflectance.pdf

However, if you cannot find the way to convert DN to TOA reflectance, just simply use DN value.

Can you specify which satellite images you are using? Can you also specify whether your research has a single or multi-temporal approach because, as mentioned before by the other users, that gives you an idea of the methodology that you have to follow

Anyway, to answer your question..I have to mention that there are 3 types of image enhancement: 1) spatial/geometric (filters), 2) radiometric (contrast stretching) and 3) spectral enhancements (ratios, indices, transforms). All in all, NDVI is a form of spectral enhancement and, generally, you do not have to apply any enhancement prior to applying it. Only pre-processing is required but I think you need to give the details that I mentioned before in order to receive the correct answer.

Hi,

First of all, you have to convert you DN to radiance and then to TOA (Top Of Atmosphere) reflectance or BOA reflectance (Bottom Of Atmosphere). Concerning BOA reflectance you have to perform atmospheric correction.

There is many algorithms can do that. You can use DOS method for getting BOT Reflectance, because is simple and effective. It's already implemented in ENVI but you have to be careful I don't recommend the automatic procedure (especially for choosing DH haze). If you want a more sophisticated and more robust approach, you have to work with RTC algorithms (Radiative Transfer Code). There is an algorithm called ATCOR implemented in ERDAS and onther one called FLAASH in ENVI that incorporates MODTRAN4 model.The problem with those algorithms is they need in-situ data such as CO2, H2O, etc. But if you don't have in-situ data you can use some defined input parameters in MODTRAN, e.g. FLAASH.

Hope that can help.

Good luck,

Adnane

I am using Landsat TM and ETM+ from 1990, 2001 and 2009. These are downloaded from USGS Earthexplorer site. If I run NDVI through Erdas Model on the raw images then what type of error arises ??

Working with with raw images is working with DN and not the reflectance, wich is wrong and will not give the real NDVI. In addition, as said before you have to perform atmospheric correction in order to mitigate atmospheric effects; especially in your case because you want to conduct a time series analysis.

Good luck,

Adnane

Gabriela Mihaela 

I have using Landsat TM, ETM+ data from USGS...to find NDVI 

You should convert  DNs to Top of Atmosphere (TOA) reflectance.