HDR using GIMP

[raaj.cee]

I've been using GIMP (on Windows XP) for a while now. I would like to know if there are any plugins/scripts available that allows combining multiple images into an HDR image. I tried searching for it but I could only find some links that mention how it's done using layers/masking. Any help is appreciated, thanks.

 

[pongheng]

I am no expert in this, but if I remember correctly, GIMP can only manipulate 8 bits images (i.e. 8 bits per channel). As such, it can not really handle HDR images.

 

[emlee]

I have been using photoshop CS (with no HDR function) to perform blending using layers on jpg files (8bit). I know GIMP can handle layers, if it can also do selection of highlights, mid tones or shadows, and can do masking and gaussion blur, then it can do blending.

I have a copy of GIMP but hasn't had the time to experiment, this is 1 thing I intend to try.

 

[osocan]

Yes there is script available. Try google for exposure blend.

I have played with it a few times but I am not into HDR to say whether it's good enough or not.

 

[raaj.cee]

Thanks guys for the replies.

I am no expert in this, but if I remember correctly, GIMP can only manipulate 8 bits images (i.e. 8 bits per channel). As such, it can not really handle HDR images.

I didn't realise that GIMP handles only 8-bit images. So true HDR may not be a possibility with GIMP.

Yes there is script available. Try google for exposure blend.

I have played with it a few times but I am not into HDR to say whether it's good enough or not.

Will look for it and experiment with it

 

[pongheng]

I think there is a confusion here - HDR = High Dynamic Range images. There are two steps involved here - first combining multiple images at difference exposure to create one image with extended dynamic range. By this, it means each channel will have 16 or more bits information from all the components images. As the traditional media can not display such images, it will have to go through the step of tone mapping - which is to map the 16 bits or more back into 8 bits. The scripts that you were referring to are to do this step. During this mapping, you define regions where you want to simply discard the lower bits information or to do some transformation to show the detailed lower bits information. Well, these are the technical details of HDR processing, you may not want to know too much about it. But, in general HDR images refer to the 16 bits or more per channel images created from the component images and not the tone mapped results.

 

[raaj.cee]

I think there is a confusion here - HDR = High Dynamic Range images. There are two steps involved here - first combining multiple images at difference exposure to create one image with extended dynamic range. By this, it means each channel will have 16 or more bits information from all the components images. As the traditional media can not display such images, it will have to go through the step of tone mapping - which is to map the 16 bits or more back into 8 bits. The scripts that you were referring to are to do this step. During this mapping, you define regions where you want to simply discard the lower bits information or to do some transformation to show the detailed lower bits information. Well, these are the technical details of HDR processing, you may not want to know too much about it. But, in general HDR images refer to the 16 bits or more per channel images created from the component images and not the tone mapped results.

If I have understood correctly, HDR refers to images that have tonal range that cannot be captured in a single image by the sensor. Conesquently, the images require 16 bits or more per channel to hold all the information. Tone mapping reduces this to 8-bits by losing some detail (choosing which detail to omit is what is done by the scripts), and is the important step in getting an HDR(exposure blended?) image. Now, since conventional media cannot display true HDR images, why go through all the trouble when tone mapping with three or more images with different exposures would suffice?

 

[pongheng]

The 16 bits or more per channel image represent the "truth" image, while the down sampled images represent only some view points of this "truth" and there are many possible tone mappings each highlight difference aspects of the "truth". This process basically mimics how human percieves the environment - difference people visiting the same place will see difference things even if they were there same place same time. Some may focused on the sky and see the patterns in the clouds, while other may focused on the trees and see the various tone of greens interplayed with the sun light. But, all of these came from the same "truth" - just difference projections. So, HDR processing attempt to recreate the "truth" by combining multiple exposures and from there create the projections that you want.

There are many mathematical details involved, and if you believe in the math, the intermediate 16 bits or more image is necessary for better processing. Which is why you can only do so much with GIMP - since GIMP can't handle this intermediate image.

 

[raaj.cee]

Wow! Brilliant explanation!!! Since the dynamic range of the final image depends on human perception as well as the ability of the display, tone mapping becomes necessary.
I did a search for the dynamic range of the human eye and got this from wikipedia http://en.wikipedia.org/wiki/Eye#Dynamic_range, http://en.wikipedia.org/wiki/Dynamic_range#Photography and http://www.clarkvision.com/imagedetail/eye-resolution.html
Exposure blending simulates the final tone mapped image and in fact reduces the tonal range in the final image so as to bring the details in both the light and dark areas, if I've understood correctly. Thanks again for the explanation!