In converting DN values to radiance as well as reflectance, do we use the same equations in raw satellite imagery for pansharpened products? I saw this DigitalGlobe technical note for DN to radiance conversion where "the information applies to all Quickbird products EXCEPT pansharpened" imagery. So may I know then how pansharpened imagery are atmospherically corrected, i.e from DN to radiance to reflectance?

  • Maybe you could rephrase the title. The questions are about converting from DN to Top of Atmosphere Radiance/Reflectance. – Nikos Alexandris Jun 9 '15 at 7:14
  • Maybe this paper from DigitalGlobe can help: ieeexplore.ieee.org/ielx7/36/6819879/… – user68132 Feb 27 '16 at 4:57

There are multiple things to respond to in your question.

1. The title of your question refers to atmospheric correction, while the information you refer to in the text is focused around converting to radiance. These two things should not be confused.

2. It does not make physical sense to atmospherically correct pansharpend data using standard methods due to the addition of the panchromatic signal to the multispectral data. However, that doesn't mean that you can't ignore those problems and do it anyway. If you choose to atmospherically correct pansharpend data you should keep in mind that the pansharpening will have changed the DN-values slightly and as such, you will have introduced a small source of errors into your reflectance errors.
All in all, my own experience shows that the errors caused by atmospherically correcting pansharpend data are minor and if the objects that you are mapping truely require the improved resolution the errors introduced are small enough to justify it.

On a related note, don't forego the chance to compare the results (a) atmospheric correction followed by pansharpening and (b) pansharpening followed by atmospheric correction. This will help you understand which errors your methodology introduces.

If your data is pansharpend by DigitalGlobe, you will need to use the formulas from the User Guide with the variable values from the supplied .IMD-file to convert to TOA Radiance. In this case, you may end up with larger than necessary errors, as you have no control of the pansharpening process. This is one of the reasons why the bundle containing separate multispectral and panchromatic bands should be purchased by technical users, rather than already pansharpend data.
Related to all of this, consider checking with your data distributor if you can acquire atmospherically corrected data from them. DigitalGlobe have recently started offering their own atmospheric correction through some resellers at a slight increase in price.

| improve this answer | |
  • Thanks Mikkel. I could not find any guide from DigitalGlobe on how to convert DN values to radiance and then reflectance for my pansharpened imagery which are Ortho-Ready Standard 2A. Funny thing is that the equations appeared to work for the 2006 Quickbird image with a clearer image and a range of positive values but for the 2010 GeoEye-1 and 2014 Worldview-3 imagery, the DN-to-radiance-to-reflectance computations produced negative values. Perhaps a lesson learned to purchase the bundle rather than pansharpened but could there possibly any way to get the reflectance values for the two images? – bean Jun 5 '15 at 2:19
  • @bean, going from DN to radiance is done using the information from the IMD-file combined with the formulas in digitalglobe.com/sites/default/files/… From radiance to reflectance requires specialized software, such as FLAASH in the ENVI software or an implementation of 6S. – Mikkel Lydholm Rasmussen Jun 6 '15 at 18:13
  • In this related question thread I posed: gis.stackexchange.com/questions/148876/…, there is a suggestion that perhaps, my data have already been converted to radiance from DN hence I don't need to do this preprocessing but go directly to atmosheric correction. As I have said, I bought images that were already pansharpened, radiometrically corrected. In this case, do they need to convert the DN to radiance for pansharpening purposes or is this a requirement for ortho-ready products? – bean Jun 7 '15 at 19:00
  • What are your minimum and maximum values, @bean ? – Mikkel Lydholm Rasmussen Jun 7 '15 at 19:10
  • Mikkel, band 1-4 (Quickbird: 0-1430; 0-1863; 0-1463; 0-1879) (GeoEye-1: all bands 0-2047) (Worldview-3: 0-1861; 0-2047; 0-1901; 0-2047). I guess these are DN then since radiance or reflectance are supposed to range from 0-1? – bean Jun 7 '15 at 22:07

you can perform atmospheric corrections on the MS and PAN bands, then execute pansharpening. It will yield better results that way

| improve this answer | |
  • Corse, the problem is I bought images that have already been pansharpened by an authorized reseller. Do you think the images have already been converted to radiance in this process? The DigitalGlobe data are Ortho-Ready Standard 2A. – bean Jun 7 '15 at 19:04
  • usually products are corrected to reflectance instead of radiance values. Then again, couldn't you drop them an email with that very question? I'm sure they would be more than willing to clarify – Corse Jun 8 '15 at 0:45

becous of changing pixels location and values in registering process, you cant perform any radiometric correction, same as radiometric correction. in first level of image correction, you should perform radiometric correction and then geometric correction.

| improve this answer | |
  • 1
    Pansharpening has nothing to do with geometric correction as long as the panchromatic data comes from the same source, which is true for most cases including this one. – Jan Šimbera Jun 6 '15 at 19:53

Not the answer you're looking for? Browse other questions tagged or ask your own question.