JimH123 Loc: Morgan Hill, CA

I have given this some thought and also tried a google search. I did find a paper written and downloadable as a PDF with a ton of math and some examples, which by the way, the merged image far exceeded the original color or original IR shot. But what was not mentioned was how to go about this. From what I saw, the color of the image remains but certain detail really stands out.

I haven't tried this yet, but I am not sure if I can take a color image and an IR image of the same scene and to blend them together like I would an HDR image and get something worthy of looking at. Or perhaps, Photoshop is needed and to process the IR in some manner that lives in one layer and to blend it with the color image living in another layer.

My google terms were "hdr merging ir and color image" and the only useful thing I found was that one paper, but it didn't include the "How".

I have a camera that is modded to be a full spectrum camera. And I have a UV-IR cut filter which can cause it to take normal color, so this is probably better than using two cameras aimed at the same scene, one IR and one color since it would be hard to have every pixel match exactly when layering.

Any thoughts?

rmalarz Loc: Tempe, Arizona

Any thoughts?
Yeah, why not give it a try and post the results.
--Bob

JimH123 Loc: Morgan Hill, CA

Yes, I want to give it a try. But understanding how to merge is another story. In the paper, the IR image contained no color information, thus it only has detail and intensity information. This can be accomplished by using a long enough wave length filter so that the red component is gone. And then this information needs to be merged with the color image, retaining the same colors and imparting to the color image the added detail.

Still pondering how to approach this. I don't think I can just use Photomatix? Or maybe I can?

rmalarz Loc: Tempe, Arizona

My first choice in merging two images would be each on a layer in PS and then change the merge mode. See if one is better as the base layer then the other. Looks like a good deal of fun exploration and discovery.
--Bob

JimH123 Loc: Morgan Hill, CA

I would think having the color layer on the bottom since you would not want to destroy the color information. Perhaps "Hard Light" or one of its neighbors since I see what it can do for sharpening.

Thanks

JimH123 Loc: Morgan Hill, CA

I just tried it. And I have to say it really works. This was a crude experiment. After getting home from work, I got out the IR modified camera, a Sony A55 which has been modded to be a full spectrum camera. And I located a spot in my yard that I could see something distant with some amount of haze present. I used a Minolta 50mm f2.8 macro lens which works good on this camera, and for that matter on any other body I try it on.

I have 3 images here:

1) The full color image. Perhaps the color could be better. I placed the UV-IR cut filter which makes the camera operate like a normal camera. But I did not do a custom WB before this image was taken since I was racing the sun which was getting lower.

2) An image taken with a 850nm IR filter. I wanted to remove all color from the image. But it was still a bit reddish pink, so I converted to B&W in Phototshop while it was the top layer of the stack. Sharpened it a bit and lowered its brightness a bit too. I am showing the actual color of the image here. If I did not convert it to B&W, the pink color bled through.

3) This is the merged image. It suffers a bit from movement of the branches of the trees. But I was more interested in the distant part of the image. In Photoshop, I auto-aligned the two layers, and then merged them using Overlay. I tried Soft Light and Hard Light, but they seemed too harsh and I went back to Overlay.

I definitely see that the color was maintained and the contrast improved. The detail in the distance really picked up as the IR image was able to cut through the haze without any issues at all. If you look at the green vegetation in the distance, you can definitely see greater detail. The farmhouse is about 3.5 miles away using Wikimapia to measure the distance.

Now of course, this experiment could have been done more carefully and to have been done with a better target, but I was anxious to give this a try and this was the target from my backyard that was available this late in the day. And by the way, I am aimed east with the sun behind me and a bit to my left. This method holds great promise when done with more spectacular scenes than this one. I was surprised how little information I found on this method when I googled. I just don't think there are many who have tried it.

Next, I will need to try this with a longer lens. Since I only have the UV-IR cutout filter in 55mm, I could still use the Minolta Beercan, the 70-200mm f4.

(Download)


(Download)


(Download)

Bobspez Loc: Southern NJ, USA

The merge really brings out the detail of the houses on the hill. On the attached I did a simple photomerge in Photoshop of your images 1 and 2 and tweaked the shadows and highlights. But your process yielded a better image and more detail of the roof tiles and above.

(Download)

JimH123 Loc: Morgan Hill, CA

Astrophotography has been using a technique like this for a long time to bring out details. From this experiment, I see that colors can be maintained while detail is added.

Opens the door to doing certain mountain scenes where the distant peaks get hidden in the haze. 850nm was the longest filter wave length I had. I suppose going longer would work even better. By the way, using the cheapee IR filters off eBay.

Need to get larger diameter filters to do this with longer lenses.