How does Atmospheric Correction work?

[ecastro]

I have been reading about it a lot, but I still cannot grasp the concept and procedure of Atmospheric Correction. If, for example, a have a raw image from a satellite, how does Atmospheric Correction do its processes, and what would be the expected resulting image?

 

[CWatters]

Not my field but...

Data received by a camera can be affected by anything that is between the object being photographed and the camera. So photos of divers taken underwater make their faces look a funny colour depending on what is in the water. Likewise data collected by a satellite is affected by what's in the atmosphere. For example if you want to know the spectra of a rock to work out it's composition that might be affected by the spectra of the light source and what's in the atmosphere between the rock and the camera/sensor.

NASA explains..

http://mars.nasa.gov/mer/mission/ spacecraft _instru_calibr.html

When you adjust the color on your television set, you do so by picking something on the screen that you know should be a certain color (such as grass should be green) and you adjust your set accordingly. This is a form of calibration. You used the color of the grass as a reference point. Instruments that go to Mars also need to be calibrated so that scientists receive accurate information. There has to be a known reference -- a calibration target...snip...They use the colored blocks in the corners of the sundial to calibrate the color in images of the martian landscape.

I imagine there are a large number of different atmospheric corrections that can be applied depending on what you are trying to achieve.

 

[Andy Resnick]

I have been reading about it a lot, but I still cannot grasp the concept and procedure of Atmospheric Correction. If, for example, a have a raw image from a satellite, how does Atmospheric Correction do its processes, and what would be the expected resulting image?

Are you referring to the process of correcting the effects of clear air turbulence on imaging?

https://www.crcpress.com/product/isbn/9780849337871

 

[ecastro]

@CWatters - Thank you for the analogy. So, if I were to do a program regarding this, should it be pixel by pixel, as long as I have a reference pixel or color?

@Andy Resnick - Yes, it is correcting the effects of the atmosphere on the image. Not just the turbulence, though. Here is an article I just read: http://www.researchgate.net/profile/Andreas_Mueller15/publication/225018984_An_automatic_atmospheric_correction_algorithm_for_visibleNIR_imagery/links/00b7d5267e226938b3000000.pdf

 

[rumborak]

If I remember correctly, one easy way of correcting for turbulence is by using a reference point that you know what it looks like and where it should be on the sky. The deformations on that reference point are then used for deriving a back-transformation for the rest of the image.

 

[CWatters]

@CWatters - Thank you for the analogy. So, if I were to do a program regarding this, should it be pixel by pixel, as long as I have a reference pixel or color?

@Andy Resnick - Yes, it is correcting the effects of the atmosphere on the image. Not just the turbulence, though. Here is an article I just read: http://www.researchgate.net/profile/Andreas_Mueller15/publication/225018984_An_automatic_atmospheric_correction_algorithm_for_visibleNIR_imagery/links/00b7d5267e226938b3000000.pdf

Perhaps google the dark pixel technique. From what I can see technique looks for dark pixels and assumes they should be black. If they aren't they assume all pixels are too bright by a similar amount...although I suspect it's far more complicated than that.

http://www.cesbio.ups-tlse.fr/multitemp/?tag=atmospheric-correction

Dark pixel method

When the image includes a surface whose surface reflectance is nearly zero, the reflectance observed at the top of the atmosphere becomes ρTOA = ρatm. We can therefore deduce the atmospheric reflectance and using a radiative transfer model, the aerosols optical thickness (AOT). Finally, knowing the AOT, we can compute the diffuse transmission, and finally calculate ρsurf. An even simpler and more approximate version of this method consists in subtracting directly the reflectance of the dark pixel (or ρatm) to the entire image (neglecting the transmission) [Chavez, 1988]. Continues..

 

[Andy Resnick]

Perhaps google the dark pixel technique. From what I can see technique looks for dark pixels and assumes they should be black. If they aren't they assume all pixels are too bright by a similar amount...although I suspect it's far more complicated than that.

http://www.cesbio.ups-tlse.fr/multitemp/?tag=atmospheric-correction

It depends on what the OP means by 'atmospheric correction'. Rumborak is correct, often a 'guide star' is used with a Shack-Hartmann wavefront sensor to correct for turbulence induced aberrations, and AFAIK, some sort of reflectance standard is often used for photogrammetric correction of imagery as well.

 

[Drakkith]

Image processing and correction is extremely complicated. My book on image processing for amateur astronomers is around 600 pages long!

 

[ecastro]

@rumborak , @CWatters, @Andy Resnick - Thank you! At least, I am confident that I need some sort of "reference" for the correction.

@Drakkith - What book is this? I wonder if it is helpful for Earth observations from space or from a satellite.

 

[Drakkith]

@Drakkith - What book is this? I wonder if it is helpful for Earth observations from space or from a satellite.

It's bundled with an image processing software, link here: http://www.willbell.com/aip/Index.htm
It's focused on observing the skies, but you may be able to get a better understanding of image processing techniques using it.

 

[ecastro]

@Drakkith - Thank you!

 

[Drakkith]

@Drakkith - Thank you!

No problem.

By the way, there's no need to tag people in your post to reply to them. You can use the 'Reply' feature to quote someone's entire post, use the 'Quote' feature to add multiple posts to the quote que (just click Insert Quotes at the bottom left of the reply box to put the posts in the quote que into the reply box), or highlight the part of the post you want to reply to and click 'Reply' or 'Quote'. In addition, replying to a post automatically notifies anyone who has responded in thread that there's a new post.

 

[ecastro]

Thanks for the advice. I didn't know that it's actually "tagging" people. I honestly thought that it is just a way to clear up to whom is the reply is for.

 

[Drakkith]

Thanks for the advice. I didn't know that it's actually "tagging" people. I honestly thought that it is just a way to clear up to whom is the reply is for.

Nope, it tags them and they get a notification that they've been tagged in a post. Just like this: @ecastro