Jump to content
The Education Forum

Recommended Posts

Sandy:

a good piece of searching on your part. Matlab is my daily bread, however, I did not know about the blind deblurring algorithm function in Matlab. I will give it a try and post any result or experience with it.

Link to post
Share on other sites
  • Replies 233
  • Created
  • Last Reply

Top Posters In This Topic

Mathias:

in principle, you could do the single-image super-resolution enhancement under Matlab on a standard laptop. However, you may need to run it for days if you would also want to do the training. Having Nvidia graphics card would be an advantage because Matlab, similar to Dr. Rudin's code, also deploys the computations to the graphics card processor. It may be better to upload their pretrained network and only enter a Prayer Man image to test the network. It may work but it may not and then you would start wondering if the training could be a culprit and would return to the training phase anyway (like using only face images in training phase if Prayer Man's face is being tested). Please let me know if you want to give it a try, maybe I can help.

Edited by Andrej Stancak
Link to post
Share on other sites

Sandy:

I have made a few tests with the blind deconvolution method in Matlab which method you mentioned a couple days ago. Here is the typical result I receive after applying the method to Prayer Man. Prayer Man image was resized from 25x25 pixels to 600x600 using bicubic interpolation. However, I also tried simple interpolation (increasing number of pixels) of different sizes as well.

The example below shows the input image (Prayer Man from Darnell after bicubic interpolation) (left panel), the recovered, deblurred image obtained with 15 iterations (middle), and the point-spread function which the program designed for this image. If I increase the number of iterations, the recovered image becomes very distorted.

 

blinddeconv.jpg

The results are not great. I cannot see more details on Prayer Man in the deblurred image compared to the original image. One useful aspect is that the Prayer Man shoulder appears to be leaner in reconstructed compared to the original image, possibly due to suppression of some reflections on the glass window in the background. I see such narrow left shoulder in my experiments with Tom Wilson's method. I will need to make changes in my 3D model of Prayer Man because I was actually modelling the wider shoulder and upper arm.

Taken together, this interesting option may not be that useful in case of Darnell's image, however, it may be useful for some other images. Thanks for pointing to the blind deconvolution method, I was able to learn something.

 

 

 

Link to post
Share on other sites
4 hours ago, Andrej Stancak said:

Sandy:

I have made a few tests with the blind deconvolution method in Matlab which method you mentioned a couple days ago. Here is the typical result I receive after applying the method to Prayer Man. Prayer Man image was resized from 25x25 pixels to 600x600 using bicubic interpolation. However, I also tried simple interpolation (increasing number of pixels) of different sizes as well.

The example below shows the input image (Prayer Man from Darnell after bicubic interpolation) (left panel), the recovered, deblurred image obtained with 15 iterations (middle), and the point-spread function which the program designed for this image. If I increase the number of iterations, the recovered image becomes very distorted.

 

blinddeconv.jpg

The results are not great. I cannot see more details on Prayer Man in the deblurred image compared to the original image. One useful aspect is that the Prayer Man shoulder appears to be leaner in reconstructed compared to the original image, possibly due to suppression of some reflections on the glass window in the background. I see such narrow left shoulder in my experiments with Tom Wilson's method. I will need to make changes in my 3D model of Prayer Man because I was actually modelling the wider shoulder and upper arm.

Taken together, this interesting option may not be that useful in case of Darnell's image, however, it may be useful for some other images. Thanks for pointing to the blind deconvolution method, I was able to learn something.

 

Andrej,

Thank you for doing that work and for reporting on it. It looks interesting though the results are disappointing.

I should have told you that you shouldn't crop the image too close to the subject (Oswald) because of the well known "edge effects" of image processing. Remember, a blurred image is equivalent to a sharp image being convolved with the point spread function (PSF). The PSF spreads each pixel of the image, thus making the resulting blurred image BIGGER than the original sharp image. By cropping too close to the subject, you are not only deleting information that could have been recovered, but you may be confusing the blind deconvolution algorithm.

Having said that, there is something about this particular algorithm that troubles me. Recall that in that instruction sheet on creating the initial PSF, it warned that the algorithm would give poor results if the initial PSF was either smaller or larger than the PSF that ceated the blurring in the first place. This raised a red flag in my mind because natural PSFs don't have well-defined sizes. Instead their edges tend to approach zero (blackness) asymptotically.

But if we pretend to be dealing with a certain-sized PSF, I can tell you one thing: Your image of the recovered PSF has an awful lot of blackness around the white part. According to the instruction sheet for this technique, this means that you chose an initial PSF that is too large to obtain a good result from the technique. (I'm, assuming that the recovered PSF that you posted is the same size as your initial PSF.)

In summary, if you are so inclined, you might want to try this again but without cropping so close to the subject, and with trying smaller initial PSFs of various sizes.

But I must say, if the original scan -- as cropped -- is only 25 pixels square, that is REALLY, REALLY low in resolution. So low that it seems to me that there really is no hope of recovering any image detail.

On the other hand, I have to question if that image really is only 25 pixels square. I mean, I can make out Ozzie's hairline, his right ear, his neck, and unbuttoned shirt. And it looks like he either has a short sleeved shirt or long sleeves with them rolled up. Could I really make those things out on a 25 x 25 pixel image?

 

Link to post
Share on other sites

Sandy:

please post or send me the piece of Darnell you would like to test and I will work on it.

The image resolution is an important aspect in any analysis, however, I guess, some details can be seen in spite of comparatively low resolution. It is as if Prayer Man's figure would be sampled with a grid of about 1 cm or slightly less.  Some very fine features will not be seen but aspects such as hairline can be seen even with this low resolution. A good interpolation during resampling does not add any new information to the image but helps to perceive the contours of known objects because our mind understands that there should be a linear continuum between two grid points on an object.

Here is the frame I used to cut Prayer Man's figure. The size of the image is 654 x 450 pixels which is a fairly sized photograph. Please indicate the region you would like to use in the analysis and I can cut it from a TIF version of this image (for Internet posting, most platforms ask for a JPEG which entails some loss of information.)

20130908-003704_1_2_1_1_1.jpg

 

Edited by Andrej Stancak
Link to post
Share on other sites
12 hours ago, Sandy Larsen said:

I mean, I can make out Ozzie's hairline, his right ear, his neck, and unbuttoned shirt. And it looks like he either has a short sleeved shirt or long sleeves with them rolled up. Could I really make those things out on a 25 x 25 pixel image?

No, actually, you can't make those things out on a 25 X 25 pixel image. For one, you can't make ANYTHING out on an image of that size. For another, you certainly can't "make out Ozzie's hairline, his right ear, his neck and unbuttoned shirt," because... it's not him.

Link to post
Share on other sites
1 hour ago, Andrej Stancak said:

And how do you know Prayer Man is not Lee Oswald?

 

Believe me, Andrej.. I would LOVE for it to be him. It would clear up SO MUCH confusion (and BTW, I admire the work you are doing with these images). But it defies logic that A) not one single person said they saw Oswald watching the motorcade from that spot and B) that he did not scream from the heavens at every possible opportunity that he was, in fact, watching the motorcade from that spot and as such could not have shot JFK.

Link to post
Share on other sites

 

15 hours ago, Jonathan Cohen said:
17 hours ago, Andrej Stancak said:

And how do you know Prayer Man is not Lee Oswald?

 

Believe me, Andrej.. I would LOVE for it to be him. It would clear up SO MUCH confusion (and BTW, I admire the work you are doing with these images). But it defies logic that A) not one single person said they saw Oswald watching the motorcade from that spot....

 

There are reasons to believe that Billy Lovelady, Bill Shelley, and Wesley Frazier all saw Oswald out on he steps and were pressured into keeping it a secret.

 

15 hours ago, Jonathan Cohen said:

....and B) that he did not scream from the heavens at every possible opportunity that he was, in fact, watching the motorcade from that spot and as such could not have shot JFK.

 

Oswald said he was out on the steps watching the presidential parade in his interrogation. An alibi so threatening to the authorities that they covered it up. This is a proven fact.

 

Link to post
Share on other sites
20 hours ago, Jonathan Cohen said:
On 9/1/2020 at 7:39 PM, Sandy Larsen said:

I mean, I can make out Ozzie's hairline, his right ear, his neck, and unbuttoned shirt. And it looks like he either has a short sleeved shirt or long sleeves with them rolled up. Could I really make those things out on a 25 x 25 pixel image?

No, actually, you can't make those things out on a 25 X 25 pixel image. For one, you can't make ANYTHING out on an image of that size.

 

I can make out all those things on that image. Just because you can't doesn't mean nobody else can either.

 

Link to post
Share on other sites
1 hour ago, Sandy Larsen said:

 

 

There are reasons to believe that Billy Lovelady, Bill Shelley, and Wesley Frazier all saw Oswald out on he steps and were pressured into keeping it a secret.

 

 

Oswald said he was out on the steps watching the presidential parade in his interrogation. An alibi so threatening to the authorities that they covered it up. This is a proven fact.

 

Which Oswald was on the steps and which Oswald was upstairs?  Which Oswald was on Elm Street filming the parade according to Martin?

3 Oswalds is way beyond my pay grade.  But, there is the problem of the Doorway Man figure.  Who is that?  It's not Billy Lovelady and whoever it is is covered by a Lovelady face mask.  

So, the Oswald on the steps must be Harvey who claimed be was outside with Bill Shelley.  Lee is supposed to be upstairs at the 6th floor and later is met on the 3rd or 4th floor by Truly and Baker or is it at the 2nd floor break room?  So, who is the Oswald figure on Elm Street in front of the TSBD by the trees shown in the Martin film?

Anyway you look at it, I can make 3 Oswalds work.  The Oswald figure on Elm Street (Martin film) could go over and become Prayer Man before Baker gets there.  Or, Harvey (Doorway Man) becomes Prayer Man. 

It is pretty much certain that two Oswalds left the TSBD by different routes.  Some folks argue differently. 

So, we have to work with two Oswalds.  One other comment here.  Harvey is a CIA man which most agree on without too much of a problem.   Can any CIA agent be trusted to be truthful?     

I have speculated from time to time there might be more than two Oswalds other than doubles.  There is really no good evidence to support the idea.  

 

Edited by John Butler
Link to post
Share on other sites

The potential of Darnell stills to unequivocally determine the identity of Prayer Man is, of course, limited by the amount of information the stills contain. However, the stills can tell us a lot about Prayer Man to narrow down the range of candidate persons or exclude them outright (such as Sarah Stanton with her thick blonde hair which just does not match Prayer Man's hair). 

The situation with Prayer Man's figure is similar to that of the figure of Buell Wesley Frazier in the same stills. In the image below, while details of Frazier's face in Darnell are not too conspicuous, we would still be able to refute Frazier as a candidate person because the image gives some good cues: the dark colour of hair, the body height, the shape of hairline and the polo type of shirt. Even in absence of details on face of Darnell's Frazier, it is possible with a very high probability to accept the possibility of Frazier being the man depicted in Darnell. His testimonies in which he placed himself into the shadow of the top landing in the western part of the doorway contribute to the acceptance of the possibility of Buell Frazier being the man there.

bwf.jpg

 

The situation regarding identification of Buell Wesley Frazier in Darnell is not so different to Prayer Man' situation. There is little chance to retrieve Prayer Man's facial features, however, the information contained in Darnell allow to narrow down the range of candidates and refute some unlikely candidates outright. 

What information regarding Prayer Man is possible to be retrieved positively and objectively from Darnell?

  • the dark tone of Prayer Man's hair suggesting dark brown or black colour of that person's hair.
  • the shape of the hairline. The hairline shows the Type II male baldness.
  • Prayer Man's body height.
  • Prayer Man's male gender.
  • Prayer Man's race of White Caucasian.
  • Prayer Man's posture (with left feet forward and weight resting on the right foot).
  • the type of Prayer Man's shirt - it was not a polo shirt but Prayer Man's forearms are bare, therefore, the shirt sleeves were rolled up to the elbows. 
  • the shirt was open in the neck region.
  • the greyness of the shirt in Darnell's representation of Prayer Man's figure matches the grey colour of Prayer Man's pants (even if details of pants are not seen).

It is also possible to infer on the range of colours which can yield the level of grey tone of Prayer Man's shirt, and the potential colours included the light red (maroon) and green colours as the first candidate colours. This analysis awaits replication.

There may be one more feature, in my view the most important one, allowing to narrow down the candidates even more, basically to only one person. This feature refers to the distribution of large dark spots on Prayer Man's shirt which appear to match the spots on shirt CE151 (the NARA 1963 picture of this shirt rather than the version which NARA sent to Pat Speer in 2016). 

Notably, none of the features listed above contradicts the possibility of Lee Oswald being Prayer Man. If Lee Oswald was not Prayer Man, at least one of the features would be different to what we know about Lee Oswald's figure and clothing he wore on that fateful Friday.

 

Edited by Andrej Stancak
Link to post
Share on other sites
21 hours ago, Jonathan Cohen said:

Believe me, Andrej.. I would LOVE for it to be him. It would clear up SO MUCH confusion (and BTW, I admire the work you are doing with these images). But it defies logic that A) not one single person said they saw Oswald watching the motorcade from that spot and B) that he did not scream from the heavens at every possible opportunity that he was, in fact, watching the motorcade from that spot and as such could not have shot JFK.

Not one single person except Lee Harvey Oswald. And he did not "scream" about anything. He was calm and collected through the entire thing. He did say he was out front with Bill Shelley.

Edited by Andrew Prutsok
Link to post
Share on other sites

 

3 hours ago, Andrew Prutsok said:

Not one single person except Lee Harvey Oswald. And he did not "scream" about anything. He was calm and collected through the entire thing. He did say he was out front with Bill Shelley.

 

Andrew,

You do know, don't you, that FBI agent Hosty's interrogation notes reveal that Oswald said he was outside watching the presidential parade? That's a fairly recent find that Bart Kamp posted here several months ago. (Time goes by so fast for me now that I don't recall if it was posted last year or early this year.)

 

Link to post
Share on other sites

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
×
×
  • Create New...