Dr. Costella announces "combined" edit improvements

[John Dolva]

Thomas, at the moment my attention is elsewhere (the nazi connection), this is not to say what you suggest is not important, it is. Others are more qualified to deal with this. I do follow and where I can offer something will do so.

Jack, this image outlines, as I understand it, the 'genealogy' of the various versions. Corrections and greater detail much appreciated. The gif compares the slide photo and Costellas alteration.

(image)

The latest version by Costella is the most altered one.(I refer to it as "candy" because of the coloring)

The closest to the original film accessible readily is in the 1964 WC publication which should be in most countries major libraries (often in mint condition), and better scans than those WC pdf's available on the net should be possible in order to reduce artefacts.

What Costella does offer, and I think it could be significant, is a gross (largescale) position correction of the gross features, which can be helpful in correctly positioning frame features in relation to each other, which can help in better seeing if frames are missing, and relative gross movements frame to frame. As far as detail goes, in all the stretching and compressing, considerable amount of detail that was there is gone, and altered. Further, considerable amount of detail that never was there is created. It needs to be recognised for what it is and any interpretations derived from it henceforth need to be evaluated accordingly. Personally I wouldn't use it except for gross stuff, and possible indicator of finer stuff, and even then with a bucket of salt nearby.

(I also think the controversy is useful in the ongoing agitation to have true copies of the original truly turned into public domain. That would be a considerable service by the powers in a position to do so.)

Also, there is nothing "magic" about layering and rotation. Many programs readily available to anyone will have such features, not just high end like photoshop. Check out the regularly appearing freebie full versions on the newsstand PC mag section.

Edited October 8, 2006 by John Dolva

[Jack White]

John...you have missed the point of Dr. Costella's work. What he

has done is present the film AS IF IT WERE SHOT WITH A PERFECT

CAMERA WITH A PERFECT LENS.

The B&H lens was NOT PERFECT. IT HAD SEVERE PINCUSHION

DISTORTION AT THE EDGES. COSTELLA HAS REMOVED THE DISTORTION.

Measurements made on a distorted image are worthless. With the distortion

removed, it is possible to observe the images ACCURATELY. Why is that bad?

Jack

[John Dolva]

No, Jack, I don't think I miss the point here at all. I probably haven't explained myself well enough. As far as I'm concerned this bit of what I said says exactly what you are saying too.

"What Costella does offer, and I think it could be significant, is a gross (largescale) position correction of the gross features, which can be helpful in correctly positioning frame features in relation to each other, which can help in better seeing if frames are missing, and relative gross movements frame to frame." To me, this encompasses gross measuring.

The point I'm making is that in applying this algorithm or formula to the image to simulate a 'perfect' lens, data at the periphery in particular has been compressed, interpolated or whatever and in the process altered. I'd go so far (and I doubt very much that Costella would disagree) as to say just about all the data transmitted through the original real life lens MUST be altered in some way.

For measuring things like angles and gross distances, this is probably in most cases fine.

When it comes to zooming in on fine details it's not. It can't be. I'm concerned that we'll see it discussed in the future as being representative of reality, hence the 'caution'.

Further, I think my sketch re 'genealogy' needs improvements. I can see I've misused some terms like resized where I should have been saying aspect ratio etc. Also there are numerous processes of data passing through lenses et'c that I understand are important.

Anyway, the point is that you can't stretch and recompress and restretch in another direction and then circularly compress and expand, resize these farmes and not have significant data loss and data creation. Each time it takes the image further away from the original. (as well there is the issue of color saturation and spectrum spread)

Also I actually doubt very much 'perfect' is a correct representation of what the distortion correction ever could be. It's a mathematical problem that can't take into account all the little imperfections of the lens in the first place, and secondly even if the lens had been smoothly, symmetrically ground and polished, no mathematical formula could be created to perfectly represent it without some unit steps, this is an empirical fact. It can only be approximated. Whether that approximation falls below a detectable (for practical purposes) threshold is another question. Naturally one would expect Costella to publish the formula he has used, so in time those qualified to do so can review.

[John Dolva]

to illustrate here is a comparison of distortion correction using Richard Rosenman's lens distortion correction plugin both directions set to -100 (max)

It leads me to wonder if all Costella did was use it? Either way I wonder if he has choosen the wrong centre for the lens? I place it a bit to the right of his. (see the difference in horizontal) Maybe I've choosen the wrong centre?

On reflection, we may have choosen different lens sizes, and as I've just guesstimated for this quick demo, I've probably choosen a slightly smaller one. Further testing (which anyone can do if they so wish, the plugins there) should clarify. I'm more intersted in the 'idea' at the moment.

Edited October 8, 2006 by John Dolva

[Jack White]

My understanding is that Costella created his own programs

for correction of pincushion distortion.

He also worked from the panorama he and I created which

shows what Zapruder's camera SHOULD depict. I shot the

panorama in 10 degree increments with a very fine ELeitz

lens, and in stitching the panorama together, he used only

the central portions of each of my photos, assuring a flat

image.

Additionally he corrected for known aspect ratio.

As far as color saturation goes, you may be unfamiliar

with the color properties of KODACHROME. It featured

VERY SATURATED COLOR. The extant z film looks very

washed out. Costella, WORKING FROM SOME COMPARABLE

1963 Kodachrome film, brought the color saturation closer

to what Kodachrome should have looked like.

Thanks for your input on this subject.

Jack

[David G. Healy]

to illustrate here is a comparison of distortion correction using Richard Rosenman's lens distortion correction plugin both directions set to -100 (max)

It leads me to wonder if all Costella did was use it? Either way I wonder if he has choosen the wrong centre for the lens? I place it a bit to the right of his. (see the difference in horizontal) Maybe I've choosen the wrong centre?

On reflection, we may have choosen different lens sizes, and as I've just guesstimated for this quick demo, I've probably choosen a slightly smaller one. Further testing (which anyone can do if they so wish, the plugins there) should clarify. I'm more intersted in the 'idea' at the moment.

Hey John Dolva, Did Rosenman publish his formula for lens distortion correction (barrel distortion/pincushioning) ?

[John Dolva]

I don't know, David, in the sense I suspect you mean it. He did 'publish' the plugin free download, and in that sense, by reverse engineering or by looking at the code itself one can see exactly what he did should one wish to do so. So I suppose in a sense, yes, he has publicly published the exact formula.

Edited October 8, 2006 by John Dolva

[Shanet Clark]

good thread

informative