UnBlur—image deblurring software
In 1999, I happened to pick up a book in the local library documenting the photographic evidence of the JFK assassination. Looking through its pages, I realised that many of the photographs and film frames were blurred—usually because the person taking the photograph or film moved during the exposure. It got me thinking about how such blurring might be removed using computers.
I was actually surprised, at first, that this had not already been done. I was aware that there had been some deblurring work done on the early images from the flawed Hubble Space Telescope (before it was repaired), and thought that images as important to the historical record as the JFK assassination would surely have been subjected to the best algorithms that scientists could muster. But the evidence was there, in front of my eyes, that this had apparently not been done.
Ideas for performing such “deblurring” swirled around in my head. I knew all about Fourier theory and deconvolution, both from my Electrical Engineering training as well as my work in Physics. But I also knew that Fourier techniques were limited, and tended to introduce artefacts that were visually distracting. So I concentrated on methods that could be applied directly in image space, rather than Fourier space.
By 2000, some of these ideas were itching to be tried out. I had done some research on deconvolution and deblurring algorithms available generally on the Internet, and found that neither were in great abundance; this explained why no one seemed to have deblurred the photographic evidence of the JFK assassination. There were hints that such methods had been used in military applications, but, understandably, the trail often “went cold”—what there was of it in public view.
I worked on these algorithms, on and off, during 2000. By early 2001 I had a workable program, and by early September of that year I had improved it substantially. I believe that the results obtained by the program are as good as are likely to be achieved by any general deblurring algorithm.
I wrote the software in a fairly flexible way, to allow me to expand it in the future. The input and output files for the program are actually in a plain text format, which allows images of arbitrary dimensionality, and accuracy, to be specified. (Apart from the obvious image processing applications, I also had in mind, for instance, three-dimensional medical scanning.)
In practice, however, the application of the algorithm to photographic images gathered the most attention. During 2001 I applied the program to frames of the Zapruder film of the JFK assassination, and was making good progress, even though the highest-quality digital scans of the film, made in the late 1990s, were not available to researchers. Towards the end of 2001, however, I became aware of a large body of JFK researchers who believed that the Zapruder film was, in fact, a fabrication. This had grave ramifications for my deblurring work, of course, both because the deblurring of a fraudulent film would be of little intrinsic value, and also because the blurring characteristics of such a film may not be completely self-consistent, depending on how the fabrication had been carried out.
During 2002, I investigated a large number of the claims that had
been put forward about the photographic evidence, stretching my physics,
engineering and computing skills to the limit. I have found that, although some
apparent indications of forgery have turned out, on closer examination, to be
mistaken (or at least not proved), there are a substantial number of
inconsistencies and impossibilities that lead me to conclude that the
photographic evidence has, in fact, been altered. Moreover, when one looks at
the broader issue of the totality of evidence, including the medical evidence
in particular, it is clear that what is seen in the extant photographic
evidence is inconsistent with what occurred in
This, together with other commitments, has led me to postpone any further work on the UnBlur program. However, the program is fully operational, in its current form, and can be downloaded and used free of charge. It is, unfortunately, still quite difficult to “drive”; a fully simplified Windows-based version of it, for simple image deblurring, would be a nice project for the future.
The following pages provide a guide to downloading and running the UnBlur program, including an introductory tutorial. Unfortunately, I am not in a position to support the program, so it is simply available to use on an “as is” basis, although I am happy to answer simple questions by email (please do not send images or files).