Improving resolution of half-toned photographs

[MarkStarr]

I am working with some 19th Century photographs which were half-toned when they were later published in books. The original prints and negatives have been lost.

Are there any techniques available for improving the resolution of half-toned photographs? I am concerned with enlarging details from these half-toned images. When substantially enlarged, the images become discrete dots, and are unviewable.

Specifically, I am wondering about filling in the blank spaces between half-toned dots in either of two ways:

(a) every blank space in a half-toned image is surrounded by four half-tone dots. Is there software that will fill in each blank space in the half-toned image with a new dot that is a composite interpolated from information contained in the four dots that surround that specific blank space? This, it seems to me, would improve the density of colors or shades of black/gray.

(b) would it be possible to use Photoshop to create a copy of a half-toned photograph and place it in another layer; then superimpose the two layers, while slightly off-setting one so the dots in layer 2 would be located in the blank spaces of layer 1. This might result in some very slight bluriness of margins of objects, but it would be offset by the increased density of the colors or shades of gray. How would one go about making this experiment with Photoshop, for example?

Any responses to the problem of improving the resolution of half-toned photographs would be appreciated. I have read that NASA has software that improves the resolution of images they reveive from space explorations. How do they do it?

Regards,
Mark Starr

 

[jimoblak]

If it is already a halftone, why not just scan it as a high resolution 1-bit image so that you maintain that halftone?

NASA and the USAF improves the resolution of images by processing several images of the same subject. This techniques is also used with video and film if the subject is stationary. This technique will not work in your situation.

 

[MarkStarr]

Thanks Jimoblak for the reply. But unfortunately, the scan is not the problem. The limiting resolution of the half-toned image is not the resolution of the scan. It is the original resolution of dots in the half-toned photograph. When a half-tone of a photograph is made, about one-third of the information in the photodraph is discarded and replaced with blank space around the dots. It does not matter at what resolution a scan of the half-toned photograph is made; the missing information that was in the original photograph is still missing.

Photoshop has a filter called 'sharpen' that attempts to fill in the blank spaces by multiplying the half-tone dots. But because it uses a 'one-size fits all' technique that does not take into consideration the size or location of each half-tone dot, this filter only succeeds in dulling or blurring the halftone image.

What is needed is a program that will measure and analyze each and every dot in a half-tone image; and then interpolate a new dot based upon averages of each group of four dots, and finally insert that new dot in the center of each group of four dots. The process could be repeated with smaller and smaller dots. The final result would not be identical to the original photograph before it was halftoned. But it would be visually superior to a scan of half-toned photograph. Most important, such an interpolated half-toned image could be enlarged much more successfully than an unprocessed half-toned image. Because the blank spaces between the dots would be filled in, the image would not break apart into discrete dots when substantially enlarged.

As a by-product, logically such a program could also get rid of moiré patterns that usually plague scans of half-toned photographs.

My question really is: is there a program already on the market that does all of this?

 

[jimoblak]

GIGO: A rule you are unfortunately stuck with.

If you don't want to reproduce the image as it is by scanning it as a 1-bit image, your other option is to scan it at an unholy resolution like 2400 ppi, apply blur until the screen disappears, then reduce the image resolution back down to something like 300 ppi. The interpolation that you seek could be achieved this way.

There is no realistic way to successfully enlarge and extract better quality than to scan the image as an unprocessed 1-bit photo. You are doomed to image loss any other way.