Stitching scans of ULF negatives. Advice?

[Linas Kudzma]

This weekend I started experimenting with scanning some of my 8x20" negatives in two halfs with my Microtex i800 and stitching them together with the photomerge command in Photoshop CS2. Typically, I scanned at 1200 ppi to give two 8x11” 1200 ppi (16 bit) grayscale tiff files, which were each ~ 250MB. I ran into several issues.

Problem one is that Photoshop needs to convert to 8 bit to stitch and I’d prefer to keep the stitched image at 16 bit until I am done working on the image. The second problem is that I ran out of memory with the full-sized files. While merging the files, Photoshop gave a message that it was out of memory (RAM). I’m running Win XP with 2 GB of RAM and a 250 GB external (USB) drive as a scratch disc. How am I out of memory?

If I downsize the files to 600 ppi Photoshop stitches them fine. Anything bigger, and it chokes. A trial version of Panorama Factory V4 refuses to even open the big files so this software seems totally useless to me.

What are others using to stitch big files from ULF negs and is there software to stay at 16 bit until all editing is done? I’m a novice at this and need advice from others that may have a way to do this.

[Kirk Gittings]

If you can track down Sandy King, he is perhaps the most knowledgeable person that I know of on this topic.

[George Stewart]

Try PanaVue Image Assembler. I've never had a problem with LF, 16-bit files. My scanner software chokes before Image Assembler does.

[Michael Mutmansky]

Linas,

The photomerge command in PS doesn't seem to work for very large files at all. I do it manually, and I suspect Sandy does the same thing. It's easy enough to do if you have good work habits, and understand how to lock the scanner exposure so that both scans have comparable color and density.

---Michael

[robc]

1. make sure that the two scans were done without changing the scan exposure settings.
2. scan so that is a slight overlap in the scans.
3. both scans must be in the exact same orientation. (Use a straight edge to slide the print to position for second scan)

4. open both images in PS.
5. crop edges to be overlayed so that they are straight and no edge of scan artifacts exist
6. create a new blank image big enough to take complete image.
7. copy both images into two separate layers in the new image
8. adjust positioning until you get an exact overlay pixel position match (zoom in to do this).
9. then flatten image and crop to taste.

If you've done it properly then you won't be able to see the join, and even if you can see it when zoomed in, you won't see it when printed. At 1200ppi you could easily do a little touching up along the join if necessary.

[John Hennessy]

It would seem that stitching parts of what was one piece of film is light work for any software. After all, the usual chore faced by a panorama program is to find, recognize and stitch edges of different files (maybe created by different films) given the variables of angles of view, changing light and wind-blown subject.

But despite the easy assignment, it is not surprising that PS is not up to the job. Maybe Adobe just does not devote the time to writing the photomerge part of PS since there are so many other very good programs.

I am only familiar with two. Both work in 16 bit and neither has ever choked on memory. The 2 GB you have should be plenty especially at only 1200 ppi. The two programs, in order of my success with them are: “PTGui” and “Panavue.” Both are shareware and they are about 50-75 usd. The former (and better) is Danish I think and the other is Canadian.

You may need to overlap more than 1" and be sure the scanning software does not automatically re configure itself for each separate scan of the same film. Make it lock on the same curves etc for the whole set of scans.

[sanking]

I do the stitching manually in Photoshop CS with 16 bit files. OS does not do merge in 16 bit, so far as I know. Here is the protocol follow.

Position your negative on the scanner in such a way that you will scan all of it in two passes. Also, make sure that the edges of the negative run on the exact right angle to the edge of the scanner. This will make final stitching much easier.

1. As has been already stated, make sure that you lock down exposure so that the two scans will be of the same density and contrast. I do this manually by first adjusting the histogram of the first scan to exact numbers, and then note this. Then you just scan the other part of the negative using these same numbers. With most software you won't do a preview before the second scan, but this may or may not be the case with all software. But if you do a second preview, you must set the numbers back to match the histogram of the first scan. In principle, however, you would not want to do a second preview.

2. When you do finish the two scans you will have two files open in Photoshop.

3. On one of them, go to Image >Canvas Size and put in a wider size to the left or to the right, clicking on the diagram where you want the original file to be. This will take a bit of time to learn so use low resolution files to practice with. I make the width 21" for stitching 12X20 negatives.

4. Now click on the other file to make it active, and using the MOVE tool drag this file onto the second file. Don't let go until this file is on the first.

5. To help see how the two files line up, change this picture layer from Normal to Difference.

6. Move the second file until the two overlap. You may drag or use the arrow keys, and use CTRL for very fine adjustments. You may also want to use the navigator to increase the size of the image on your monitor to give you a better view.

7. When the two files match the overlapping area wil turn black, or almost so.

8. Now turn the layer back to NORMAL.

9. Soften the edges if necessary by erasing with the brush tool (set towhite), or with the eraser tool (with background color white).

10. When it looks good, flatten the layers and save the file.

11. Crop off the extra white space.

This may sound complicated but after you do it a few times you will get the trick. But remember, in order to get a perfect stitch it is very important that you have files of the same density and contrast, and that the edge of the negative runs at the same right angle to the end of the scanner.

Sandy

[Frank Petronio]

Isn't it ironic how using ULF forces these photographers to become more adept at using Photoshop?

It certainly is healthy: Prevention of Alzheimer's Disease (http://www.memory-key.com/Seniors/Alzheimers_Prevention.htm#halve).

[Linas Kudzma]

Thanks to all for the detailed answers! Locking the scanner exposure for both scans has not been a problem with the Microtek i800. I downloaded a trial version of PTGui and it does indeed stitch 16 bit files. I will also try the manual stitching Sandy and rob describe, if not for any other reason than for mental exercise to postpone dementia.

[Andy Jones]

As others have indicated, you don't need merging software beyond PS. I shot two 4x5's from one set-up using complete back-cross (+/- 60mm) to create a "4x10" image. Both were developed in the same run and scanned identically at 2500 ppi at 16 bit. While creating two large files (each ~600MB) this did not create any problems for PS except that it took a long time to open them both, increase the canvas size for one and paste the other into it. (I have 3.5 GB RAM on a dual 2GHz G5, which may account for it; however, with less RAM I'd just expect it to take longer, using the scratch disk more). Adjusting the position and (slight) rotation of one I had made 50% transparent, allowed me to align them, albeit slowly. No exposure adjustments or distortions were needed (I have a good lab to developthem!) Using a small radius feathered selection tool I selected the parts of the overlap of the left edge of the right image (where there were no detailed components of the image) to create a smooth and non-linear seam and deleted them. Flattened and processed as one file thereafter. With a slower system, dementia should not set in before the paste is complete :-)

[tim atherton]

hmm - some useful tips - I've got to stitch groups of 20-60 images together of scrolls that were photographed in sections - that should be fun :-(

[Daniel Moore]

Your first and last stop should be PTGui. Any image transformation you can possibly dream up will be supported by it. The Panotools yahoo group is a sea of knowledgable folks who can provide expert support. PTGui, PTAssembler (also top notch) Hugin, PTMac and couple other programs are interfaces for PanoTools code written by Helmut Dersch. All designed for stitching with various degrees of automation. Many support Enblend, Smartblend and Autopano. These take care of automating exposure and color compensation, and automating 'control point' generation. These programs have developed now to the point of add images, press button, panorama. A key point to aligning scanned images is to set a FOV to a very low number like 1, tricking the software to thinking it's working with images taken with a virtual telescope and to eliminate correcting optical distortion. Again, Pantools forum over at Yahoo has answers to questions you haven't thought of yet. Overlap is not usually a concern of the scannist, but needs to become one to stitch effectively. More time, but recovered with reduction post processing corrections.