## Friday, November 8, 2013

### Analysis of Pixel Diffusion In Oswald 'LIFE" Photos

Regions near right eye in 2 photos and fractional pixel deviations for Dt(xb)  and Dt(yb)

Four years ago Dartmouth prof Hany Farid presented alleged "proof" the Oswald backyard rifle photos were "real". Google the name “Hany Farid” and you will bring up about 5,000 pages on the Dartmouth computer software scientist – from almost as many news sources – all parroting that he has “proven the Oswald backyard rifle photo is genuine." (One begins to wonder from this if the CIA's "Operation Mockingbird" is still going on.). For those who may not know, this is the infamous photo taken in the Oswalds’ backyard on Neeley Street in Dallas, sometime in March, 1963 (as documented from FBI files by Mark North in his ‘Act of Treason’) .

I did my own analysis of the backyard photos, based on the reasonable assumption that any transformation from specular dots on silver iodide based (1963) photo emulsions will undergo some drift  over time in rectangular coordinates such that this time-affected diffusion needs to be examined using fractional calculus equations of the form : d’(x) = wH×(d(x)bH) and d’(y) = wV×(d(y)bV).

An illustration of applying the parameters d'(x) and d'(y) to respective pixel sub-frames is shown in the accompanying graphic. These measure fractal deviations in pixel density from one Oswald photo  to the other (there were 4 in all) and from which the weighted error in the conceptual space may be computed.

I give some of the results below for the diffusion factors, Dt(xb)  and Dt(yb) done using a Mathcad 14 software program. This is based on comparison of two photos (A and B) with the same regions selected as “prototypes” and the mapping done one to one, e.g. from points of dx1 to dx1’ and dy1 to dy1’. We start out with initial values: y = dy1 = 2.7 mm, and x = dx1 = 3.4 mm.

I estimated the weighting parameters: wH = 1.01,  wV = 1.03, bH= 0.0015,  bV= 0.10, so that: w = 1.443 and b = 0.1. With the non-dimensional fractional thickness change parameter estimated at: t = 0.001 then: Dt(xb)  = 0.967 and Dt(yb) = 0.964.

The total mapped diffusion from (A) is:

In x-coordinate:  dx1’ =  {dx1 +  Dt(xb)}   = 4.3 mm

In y-coordinate: dy1’ =  {dy1 +  Dt(yb)}   = 3.6 mm

The change in aspect ratio is: 0.794 to 0.839 or 5.6%

The original Euclidean resultant (for A) is:

DrA(x,y) =      [ (dy1)2 + (dx1)2]1/2   =

{(y2 – y1)12 + (x2 – x1)12}1/2  = 4.34 mm

The mapped (diffused) resultant (for B) is:

DrB(x,y) =      [ (dy1’)2 + (dx1’)2]1/2   =    {(y2’ – y1’)22 +           (x2’ – x1’)22}1/2  = 5.70 mm

The fractional error for the resultant is then: d(d.AB)  » ½ (f1) + ½ (f2)  where:

f1=  [DrB(x,y) - DrA(x,y)] / DrA(x,y)  = 0.31    and

f2 = [DrB(x,y) - DrA(x,y)] / DrA(x,y) =   0.23

We now look at another example for which the weighting parameters are significantly increased. Again, we use the same original dimensions: y = dy1 = 2.7 mm, and x = dx1 = 3.4 mm.

We estimate the new weighting parameters: wH = 1.05,  wV = 1.1, bH= 0.05,  bV= 0.18, so that: w = 1.521 and b = 0.187. With the non-dimensional fractional thickness change parameter estimated at: t = 0.4 then: Dt(xb)  = 1.039 and Dt(yb) = 1.014.

The total mapped (length) diffusions from (A) are:

In x-coordinate:  dx1’ =  {dx1 +  Dt(xb)}   = 4.43 mm

In y-coordinate: dy1’ =  {dy1 +  Dt(yb)}   = 3.71 mm

The change in aspect ratio is: 0.794 to 0.837 or 5.3%
The original Euclidean resultant (for A) is:

DrA(x,y) =      [ (dy1)2 + (dx1)2]1/2   =
{(y2 – y1)12 + (x2 – x1)12}1/2  = 4.34 mm

The mapped (diffused) resultant (for B) is:

DrB(x,y) =      [ (dy1’)2 + (dx1’)2]1/2   =
{(y2’ – y1’)22 + (x2’ – x1’)22}1/2  = 5.78 mm

The fractional error for the resultant is then: d(d.AB)  »
½ (f1) + ½ (f2)  where:

f1=  [DrB(x,y) - DrA(x,y)] / DrA(x,y)  = 0.33    and

f2 = [DrB(x,y) - DrA(x,y)] / DrA(x,y) =   0.25

We find that with realistic estimates of the weighting parameters, there is very little diffusion difference in the mapping of pixels from A to B along the dimensions indicated. Given some original values dx1, dy1, the mapping will yield diffused dimensions that are within about 30% of the original.  However,  +30%  shows that there clearly had to have been tampering such that the images are not the same. Note, however, the purpose of the fractional calculus diffusion method is to show that two images are not the same in terms of pixel distributions. It does not claim to show how this happened.

In addition to the above cautionary points, the deviation of optical density for solar radiation falling at an angle of incidence i exhibits only minor diffusive differences (e.g. << 5%) on such (silver iodide) film negatives. Indeed, for years when teaching introductory astronomy courses, it was precisely this property that allowed me to show my students how to use the old b&w film negatives to view the Sun during partial eclipse without risk of blindness. One can't do that with the emulsions used today - assuming one uses old fashioned cameras! Hence, Farid's counting on computer processed pixel re-transformations of old (presumed) silver iodide films today to affirm fakeness is akin to me looking at old home movies on dvd format and trying to ascertain where the exact point of editing occurred.
----

In Farid’s computer analysis, one is left to wonder what exact photo he has proven genuine- since there were four in all. One of those featured small irregularities including that the telescopic scope was absent, because a technician had accidentally retouched it. WHY has Farid not picked it up with this elite software, when it was openly admitted by the management of LIFE magazine? Or, was Farid not given the retouched photo? If not, why not? Perhaps to prevent him from saying that ONE photo at least was a fake?

Then there is the “Oswald ghost” photo recovered at Dallas PD headquarters some time after the assassination. It is shown below:
As one can discern, the “ghost” is a cutout into which another image can be pasted-superposed. The cutout image, many of us conclude, was obtained using a Dallas cop stand –in, which photo was also found in Dallas Police files, along with the ghost image.  Photo specialist Robert Hester was called on 22 November, 1963 to help process assassination -related photos for the FBI and Dallas police. Hester reported (and his wife Patricia confirmed) that he saw an FBI agent with a color transparency of one of the backyard photos with NO figure in the picture. This has to be the same as that shown above. Was the FBI in on the manipulation of images and photos? We don't know, but given Hoover hated JFK's guts and Hany Farid's connection to the FBI in funding his lab, can we really trust his work? Can we trust he analyzed the actual source photo? And if so - which?

Now, I am not implying here that Farid faked the photos used for his work, so we are clear. However, he had to have obtained the backyard photos he used from someplace and most plausibly it was from the FBI -which funded his lab along with the Justice Dept. We already are well aware of the record of the FBI’s chicanery in JFK assassination –related evidence tampering: for example as noted earlier, and also when FBI special agent Richard Harrison (equipped with finger print ink and Oswald's alleged rifle) went to Miller’s Funeral Home where the slain Oswald lay, in order to get ex post facto fingerprints which weren’t there originally. As Walt Brown notes (Brown: 1995, Treachery in Dallas, Carroll & Graf Publishers, p. 332.), the Dallas PD never bothered to print him at the time of booking - incredible as that seems-  but lots of incredible events transpired that day and on those to follow, including Oswald's killing by Jack Ruby.

In 1977, Canadian Defense Dept. Photographic specialist Maj. John Pickard noted the 99% probability the LIFE cover photo was a fake and noted each photo was taken from a slightly different angle. When superpositions of the images are performed, e.g. one photo laid atop another in succession, it is found that nothing matches exactly. As Pickard observed:

"Yet, impossibly, while one body is bigger - the heads match perfectly."

Again, this can be explained using the same "spliced out head"  re-incorporated into each image..

It is incredible Farid has not performed this simple test, but then that would mean divorcing himself from his computer- which he obviously believes to be the next thing to an Oracle. In the above context, I am convinced that Farid’s preoccupation with statistical disparities or deviations in pixel density is only useful provided he knows the full and complete history of the object-film-photographic emulsion he is investigating. In particular, what criteria has he to implicitly trust the source photo? As my middle brother, a former photographic specialist with the Air Force in the 1960s also pointed out to me

First of all he'd have to have had the original photo of Oswald, not a copy, which I doubt very much he had, since the Feds confiscated it and only allowed reprints. Secondly, in those old photos there were no pixels to measure since all photos of that period were taken with film type cameras and no pixels were on them to measure. The photos looked the same as a painted picture, smooth and even ”

To add to what he said, I do admit Farid could have generated pixels from the old photos (to enable and facilitate computer processing), but also - as I pointed out- he'd have to use fractional calculus at all inter-comparison points between all the relevant photos to be assured of a faithful reproduction or representation following any transformations of media. Since it isn't at all clear he did this, then basically he revived computer formats of the old photos, which appear to have already been tampered with, and reproduced the same errors. Farid actually re-discovered the tampering, but was too blinded by his computer processing tools to see it. To make a long story short, and skip to the chase, the differential fractional calculus transformations disclose that the V-shaped shadow under Oswald's nose is essentially identical in all photos, despite the fact the photos were taken at different times.