Creature Suit Analysis - Part 6 - Comparative Anatomy

[Guest longtabber PE]

You keep saying 1D as Gigantofootecus points out it's 2D

and ( not wanting to start a fight- I'll address that shortly) thats simply incorrect ( in context with what I'm saying- photography has a different view)

[Guest longtabber PE]

Why?

Unless the object in the photo is too close to the camera, the projection (3D->2D) doesn't distort the image. The problem is to restore the Z component from a 2D image.

The curvature of the earth only comes into play at smaller scales, typically 1:5000 or smaller from an overhead POV. We wouldn't actually have to ortho-rectify Patty's image. The exercise here would be to account for the foreshortening effects due to orientation of a 3D object projected on film and the distance from the camera.

Foreshortening effects?

This is called mono-restitution. Restoring the Z component on a 2D ortho-image. The method you describe uses 3 or more control points and space resection to rectify the image. This removes the tilt displacement of the camera, but you still require distances from the camera and the relief of the object to restore the Z component. All this does is correct for the distortion caused by the projection (ortho-rectification). But it's not a good approach for "flattening" out Patty's image since we don't have enough information to do this. We would, however, only require to determine the nodes of the skeletal structure, and not the entire image.

I think determining the range IS the exercise. This would tell us how much guesswork we're dealing with. We can't know until then.

I wouldn't assume it can't be done. Just not with the method you described. This is a job for forensic photo-metrology.

This is a 2D analysis. In this case the distortion is strictly due to foreshortening caused by the orientation of the image and the relative distance from the camera. As Jack mentioned, if the object is 100+ feet from the camera then the image is distorted at most 2-3% due to the varying depth of the object. But this applies to the scale of the visible body parts, and is only significant at distances such as Soarwing's transparent overlay. In this regard, it's actually fortunate that Patty was filmed at far range. The closer to the camera, the greater the distortive effects.

There are several modern techniques for overcoming the quality of the film, namely feature tracking and frame averaging, which are methods of motion analysis. None of these methods have been applied to the PGF because the original full frame version has not been available for study. And no photogrammetric study is going to happen with the copyright issues involved. But it certainly can be done and whether such an analysis would glean any useful information is speculation at best. Otherwise, no one can know that it wouldn't.

Which brings me to Soarwing's transparency overlay. This is an invalid comparison. Soarwing is 30 feet from the camera, which means the length of his right arm and foot board is ~7% overscaled relative to his body. At any rate, there's way too much distortion for an accurate comparison. The shots of him farther from camera are more appropriate. But too much is being made of his girth matching up with Patty's. Unmatched body angles can give the illusion that he matches up in girth, when he doesn't.

Of particular note, his arm length definitely comes up short. His leg lengths and angles are difficult to determine for a comparison since his ankles aren't visible. But note how much higher his bent left knee is relative to his right knee compared to Patty's. I don't think their legs match up well at all. Too many assumptions are based on these comparisons.

Also mentioned was how spherical aberration might effect an image. It would depend how far the image is towards the periphery of the lens. You can really see this kind of distortion on the edges of Rick Noll's blowups under the microscope. The edges are quite distorted, much more so than from a normal camera lens. Since the quality of the lens for the K-100 was decent, the only significant distortion would occur outside the sweet spot. But there were plenty of images taken near the center of the frame. And aspect ratio is a film, not a lens feature. The transfer from 1 medium to another can also affect the aspect.

As wide as Ray G's back is, his ASH ratio is still under 1.10. I found it impossible to fit his image onto the "back" shot of Patty (I posted upthread). But this was due to his head not his back. Lining up his eyes/ears and rescaling his head and shoulders to fit just didn't work. Patty's head appears smaller relative to her body than Ray's. The point here is that individual body parts may not be outside the human limits, but IMO, collectively they make for a very disproportioned individual. The question is whether all this can be disguised inside a suit? Formal analysis 1st, speculation 2nd.

Lets go by the numbers here

>>>Why?

I explained that earlier but its important to point out that the terminology differs substantially from a photographers view ( image ) than whats necessary for 3 D modeling

>>>Unless the object in the photo is too close to the camera, the projection (3D->2D) doesn't distort the image. The problem is to restore the Z component from a 2D image.

No, thats factually incorrect ( I know where you are coming from- its the difference between an isometric drawing and an ortho one) thats not the same thing as true 3D modeling. You cannot "restore" anything unless you have fixed coordinates. In this case, they dont exist. "distortion" is included unless you have a perfect lense, in perfect distance with known measurements- anything beyond that is distortion. Thats a mathmatical fact.

>>>The curvature of the earth only comes into play at smaller scales, typically 1:5000 or smaller from an overhead POV. We wouldn't actually have to ortho-rectify Patty's image. The exercise here would be to account for the foreshortening effects due to orientation of a 3D object projected on film and the distance from the camera.

Foreshortening effects?

That was nothing more than example- the condition exists regardless, the point is that without "knowns" everything is speculation ( unless one wants to try to say that all the vectors dont matter- in that case, I'll pull out the math)

>>>This is called mono-restitution. Restoring the Z component on a 2D ortho-image. The method you describe uses 3 or more control points and space resection to rectify the image. This removes the tilt displacement of the camera, but you still require distances from the camera and the relief of the object to restore the Z component. All this does is correct for the distortion caused by the projection (ortho-rectification). But it's not a good approach for "flattening" out Patty's image since we don't have enough information to do this. We would, however, only require to determine the nodes of the skeletal structure, and not the entire image.

I think determining the range IS the exercise. This would tell us how much guesswork we're dealing with. We can't know until then.[/i]

In this case, you and I are saying the same thing but we are speaking from 2 different perspectives. I am looking at micrometer precision and you arent. At the end of the day, the fact that the deduction of these figures factor greatly in the question of if patty is in a human range or not( only because this scale is in inches and important). As I bolded ( which represents what I said)- we "dont" know and thats a major part of this analysis.

>>>I wouldn't assume it can't be done. Just not with the method you described. This is a job for forensic photo-metrology.

This is a 2D analysis. In this case the distortion is strictly due to foreshortening caused by the orientation of the image and the relative distance from the camera. As Jack mentioned, if the object is 100+ feet from the camera then the image is distorted at most 2-3% due to the varying depth of the object. But this applies to the scale of the visible body parts, and is only significant at distances such as Soarwing's transparent overlay. In this regard, it's actually fortunate that Patty was filmed at far range. The closer to the camera, the greater the distortive effects.

BINGO I didnt "assume" it cant be done- I just stated because of the film it cant be done ( with any degree of accuracy) and again, we are saying the same thing- just different perspectives. Thats only important because of the close tolerances addressing the question of "can a human fit into this or not"

>>>Of particular note, his arm length definitely comes up short. His leg lengths and angles are difficult to determine for a comparison since his ankles aren't visible. But note how much higher his bent left knee is relative to his right knee compared to Patty's. I don't think their legs match up well at all. Too many assumptions are based on these comparisons.

again, all of that is subjective because the exact parameters arent known. You are doing a "picture" analysis and I'm doing a dimensional analysis- your model allows for variance, mine doesnt

>>>Also mentioned was how spherical aberration might effect an image. It would depend how far the image is towards the periphery of the lens. You can really see this kind of distortion on the edges of Rick Noll's blowups under the microscope. The edges are quite distorted, much more so than from a normal camera lens. Since the quality of the lens for the K-100 was decent, the only significant distortion would occur outside the sweet spot. But there were plenty of images taken near the center of the frame. And aspect ratio is a film, not a lens feature. The transfer from 1 medium to another can also affect the aspect.

I said that

>>>As wide as Ray G's back is, his ASH ratio is still under 1.10. I found it impossible to fit his image onto the "back" shot of Patty (I posted upthread). But this was due to his head not his back. Lining up his eyes/ears and rescaling his head and shoulders to fit just didn't work. Patty's head appears smaller relative to her body than Ray's. The point here is that individual body parts may not be outside the human limits, but IMO, collectively they make for a very disproportioned individual. The question is whether all this can be disguised inside a suit? Formal analysis 1st, speculation 2nd.

LikeI said ( and you demonstrated)- its imposible to fit an image because of the unknown

[wiiawiwb]

Giganto.... I admire your staggeringly honed acumen. I'm not sure what you do to put food on the table but your breadth and depth of knowledge is absolutely remarkable.

Edited January 28, 2008 by wiiawiwb

[Guest soarwing]

Which brings me to Soarwing's transparency overlay. This is an invalid comparison. Soarwing is 30 feet from the camera, which means the length of his right arm and foot board is ~7% overscaled relative to his body. At any rate, there's way too much distortion for an accurate comparison. The shots of him farther from camera are more appropriate. But too much is being made of his girth matching up with Patty's. Unmatched body angles can give the illusion that he matches up in girth, when he doesn't.

- - -

Giganto,

My intent was to match myself up to Patty as close as I could - Given a 14.5" footprint left by Patty and the same length plank on my foot.

I know that it's not 100% accurate.

However, the notion that Patty has huge, "inhuman" girth or is somehow very tall isn't supported by the evidence - again, given that the filmed subject has a 14.5" foot.

I was wearing only a t-shirt in some of the photos and a light jacket in some others. No additional padding of any kind.

I realize that there is some over-exposure/smearing going on in the image of Patty's foot, but not enough to change the comparison significantly. Even giving Patty the benefit of the doubt, I still struggle to understand the "inhuman" arguments.

There's just no reason to think that she was of "inhuman" size or proportions, especially considering how little I was wearing and what you see are my bare hands, my actual head and the rest of my unpadded body. If I were to do the same comparison wearing a padded "muscle suit", monkey gloves and a conical skulled headpiece, it would make Patty seem even less impressive - certainly not of inhuman size or proportions. Maybe you could make the case for Patty being of unusual proportions, but I think that "inhuman" is something of a stretch, given the obvious problems getting decent measurements of the actor potentially inside the suit.

The longer range shots I did validate the general finding - albeit rough - of the closer shots. Even adding 7% to Patty, she fails to fall within some inhuman size or height. Especially given that a suit with padding undoubtedly would have been needed for the PGF creature. The errors I made may not all go Patty's way as well.

Even though I tend to support the authenticity of the film, I think the "Patty is massive" arguments have little evidence supporting them.

[Guest soarwing]

Soarwing's pics are about the best I've seen in putting a human being inside the outline of the PGF creature. More than ever now a suit can't be ruled out. What I find especially revealing is the way the board attached to the Soarwing's foot duplicates the so called extended heel of the PGF creature. This makes great sense as extending a foot completely to the front would look completley fake. Also Patty's foot does indeed look more like slab of wood or rubber than it does a flesh and blood object. That "paddle foot" frame has bothered me from the first time I saw it. IMO its the most telling element of the creature that points to a hoax.

- - -

Thanks.

I think my neighbors were entertained at least.

The paddle foot effect is due largely to overexposure and perhaps some image smear on top of that.

Her left foot looks pretty normal in other frames, IMHO.

Edited January 28, 2008 by soarwing

[Guest]

H & V are part of a 1 D drawing ( read any blueprint- thats what you see a drawing referenced to a X and Y axis)

H & V and X & Y sure look like 2 Dimensions to me

[Guest soarwing]

To:

Lyndon, Longtabber PE, Soarwing, Killian, dogu4, Ray G, Bipedalist, Jack, Gigantofootecus,

Thank you all for contributing posts that are well reasoned, informative, civil, and constructive to everyone's understanding.

Just wanted to acknowledge all your fine contributions.

Bill

- - -

Let me just say that I've enjoyed your information and presentations tremendously. Fascinating stuff.

My efforts are primitive compared to those of the other names you mentioned, but you're most welcome.

[Guest Lyndon]

To:

Lyndon, Longtabber PE, Soarwing, Killian, dogu4, Ray G, Bipedalist, Jack, Gigantofootecus,

Thank you all for contributing posts that are well reasoned, informative, civil, and constructive to everyone's understanding.

Just wanted to acknowledge all your fine contributions.

Bill

Well I wouldn't put myself in the same category as those other fine clever lads you have mentioned but I appreciate your comments.

I'm glad to see these threads of yours pinned so they don't get lost in amongst the bickering threads (of which I have unashamedly engaged in myself).

Edited January 29, 2008 by Lyndon

[Guest]

No, thats factually incorrect ( I know where you are coming from- its the difference between an isometric drawing and an ortho one) thats not the same thing as true 3D modeling. You cannot "restore" anything unless you have fixed coordinates. In this case, they dont exist. "distortion" is included unless you have a perfect lense, in perfect distance with known measurements- anything beyond that is distortion. Thats a mathmatical fact.

I don't want to turn this into an argument, but an isometric view is hardly a true 3D model. It's a distorted projection of the true perspective of an image typically used for blueprints and other CAD views. An iso-view removes the Z component as well as the actual 3D perspective. So how does this apply to the PGF? The PGF is a collection of raster images, not vector diagrams. Given enough information, an ortho view can restore the Z component from a 2D image. An isometric transform burns it's Z bridges behind it.

I didn't mean to imply that any view doesn't include some level of distortion. Only an object at infinity with a perfect lens would not be distorted. But you still haven't explained why Patty's image would be reduced by 5%. In this case the level of distortion is dependant on the distance from the camera and the orientation of the figure. At 150+ feet from the camera (frame 72) the max distortion is ~2% if there is 3 feet of surface relief visible on the image.

The exercise here is to establish a range of error, not to create an iso-view with micrometer precision.

That was nothing more than example- the condition exists regardless, the point is that without "knowns" everything is speculation ( unless one wants to try to say that all the vectors dont matter- in that case, I'll pull out the math)

What "knowns" are you referring to and why are they unknowable? And where did I imply the vectors would not matter? Go ahead and pull out the math so I know where you're going with this. The PGF isn't a CAD drawing. This is a "picture" (raster) exercise. If we're talking about rectifying the images, then IMO a DEM (digital elevation model) would be most appropriate, except we'd need a surface model of Patty rather than an elevation model. But this puts the cart before the horse since it's the model we're trying to determine. However, it's still more feasible than somehow deriving isometric transforms to do this.

Soarwing, you acknowledged that you were too close to the camera in your transparent overlay, yet you continued to use this comparison to claim that Patty is humanly proportioned. You also claimed that an image distortion of 7% isn't significant. How exactly did you determine that your limbs matched up with Patty's? Why don't you take another photo of yourself at 180+ feet from the camera walking directly away from the camera and superimpose it over the frame I posted upthread showing Patty's full back width and let's see how your girth matches up? In particular, try to fit your head inside the suit scaled to the same height as Patty.

Here's Patty's image scaled up 7%. (click to animate) I'm not claiming this is accurate, but do you still think a 7% difference isn't significant?

Here's your arms scaled to a common length. The smaller Patty appears, the longer her arms are relative to yours (providing no prosthetics are involved). Still think you're similarly proportioned?

I wanted to do the same thing with your legs but your ankle appears to be hidden in the grass. Since I'm the same height as yourself, I assumed that our tibias are the same length. Mine comes to 17-18" (ankle to the center of knee, which doesn't have to be exact, just common). Let's use the lower estimate of 17". So if the plank on your foot is 14.5" then I can derive the placement of your ankle if your tibia is also 17". Frame 72 might not be the best for estimating Patty's tibia, but I took a stab at it anyways. If my estimates are close, and you were the person inside the suit then here's the suit you'd be wearing.

At any rate, I don't see how you can claim that Patty is humanly proportioned. Approximating the height and alignment by overlaying images isn't a particularly good measuring tool unless you match up all the variables. But we need to establish them 1st.

[Guest Apeman]

Enjoying this.... and seeing things more or less the same as GF.

And still wondering where LT's 5'8" - 5'10" came from and how exactly any "experience" was used to derive that number?

Apeman

[Drew]

At any rate, I don't see how you can claim that Patty is humanly proportioned. Approximating the height and alignment by overlaying images isn't a particularly good measuring tool unless you match up all the variables. But we need to establish them 1st.

I don't think whether or not Patty is humanly proportioned is important, if you consider that if it is a Suit, it was probably made to alter proportions of the human inside of it.

Here is one way of altering the proportions that a prominent Gorilla Suit manufacture lists as one of the ways to bulk up a Gorilla suit: "There are many ways one can bulk up a suit. A "yoke" can be made from upholstry foam. This yolk fits around the back of the neck, around the shoulders. When made correctly, this yoke will also give the illusion of the arms being longer than they really are because the yoke raises the natural contour of the shoulders. It also enhances the sagittal crest and the neck (or the lack of). If you look at adult gorillas, they have very little "back neck.""

[Bill]

Drew:

There is a problem with the "yoke" of padding you referred to. Raising the shoulders can be done to apparently lengthen the arm (excellent example, the "Metaluna Mutant" creature from This Island Earth, a 50's si-fi classic) but there's a catch.

Raising the shoulder to apparently lengthen the arm is an illusion most believable if the human keeps the arm essentially straight down. Raising the arn outward sideways, or swinging it strongly forward or backward, ruins the illusion, because it allows us to determine where the real shoulder socket is, and thus also see that padding artificially raised the shoulder.

In the PG Film, the arm swings far forward and then far backward, approximating an estimated 90 degree arc counting both directions of swing. If you take a line up the arn center (going parallel to the arm bicep and tricep mass outer edges, you can easily generate two lines that should intersect at about the real shoulder socket under the fur.

If you pad up the shoulders of a suit, that intersection of lines will tell you the shoulder socket is in an abnormally low position, and reveal a padded suit is being worn. It's one of the first things I looked for, and I personally saw no evidence of such.

So people wearing suits with padded shoulders and trying to create an appearance of realism, don't swing the arms like that if they can possibly avoid it.

Bill

[Guest longtabber PE]

I don't want to turn this into an argument, but an isometric view is hardly a true 3D model. It's a distorted projection of the true perspective of an image typically used for blueprints and other CAD views. An iso-view removes the Z component as well as the actual 3D perspective. So how does this apply to the PGF? The PGF is a collection of raster images, not vector diagrams. Given enough information, an ortho view can restore the Z component from a 2D image. An isometric transform burns it's Z bridges behind it.

I didn't mean to imply that any view doesn't include some level of distortion. Only an object at infinity with a perfect lens would not be distorted. But you still haven't explained why Patty's image would be reduced by 5%. In this case the level of distortion is dependant on the distance from the camera and the orientation of the figure. At 150+ feet from the camera (frame 72) the max distortion is ~2% if there is 3 feet of surface relief visible on the image.

The exercise here is to establish a range of error, not to create an iso-view with micrometer precision.

What "knowns" are you referring to and why are they unknowable? And where did I imply the vectors would not matter? Go ahead and pull out the math so I know where you're going with this. The PGF isn't a CAD drawing. This is a "picture" (raster) exercise. If we're talking about rectifying the images, then IMO a DEM (digital elevation model) would be most appropriate, except we'd need a surface model of Patty rather than an elevation model. But this puts the cart before the horse since it's the model we're trying to determine. However, it's still more feasible than somehow deriving isometric transforms to do this.

>>>I don't want to turn this into an argument, but an isometric view is hardly a true 3D model. It's a distorted projection of the true perspective of an image typically used for blueprints and other CAD views.

I said that- its a picture- not a drawing- the key here is distorted ( that was the whole point)

>>> An iso-view removes the Z component as well as the actual 3D perspective. So how does this apply to the PGF? The PGF is a collection of raster images, not vector diagrams. Given enough information, an ortho view can restore the Z component from a 2D image. An isometric transform burns it's Z bridges behind it.

Again, that was the whole point ( maybe I should have wordsmithed it beter)- there is no way to get precise enough information for the Z axis for accurate reconstruction. So far we are saying the same thing.

>>>I didn't mean to imply that any view doesn't include some level of distortion. Only an object at infinity with a perfect lens would not be distorted.

Again, thats the same thing- the point I was emphasizing was there is distortion and theres no way to accuratelty account for it.

>>>But you still haven't explained why Patty's image would be reduced by 5%. In this case the level of distortion is dependant on the distance from the camera and the orientation of the figure. At 150+ feet from the camera (frame 72) the max distortion is ~2% if there is 3 feet of surface relief visible on the image.

I think this is where the point was missed. You are referencing the lense distortion ( as you stated around 2%) and I am including the additional "unknown" factor of imprecise dimensions because the actual coordinates are unknown as are exact ranges- this I said between 3-5% ( which the 2 combined are in line with yours)

>>>The exercise here is to establish a range of error, not to create an iso-view with micrometer precision.

And thats what we both have done

>>>What "knowns" are you referring to and why are they unknowable? And where did I imply the vectors would not matter?

I dont recall saying you said ( or implied) vectors didnt matter ( but I was commenting on several threads and if I put that in a response to yours- that was an accidentalm oversight to not highlight that distinction)

>>>And where did I imply the vectors would not matter? Go ahead and pull out the math so I know where you're going with this. The PGF isn't a CAD drawing. This is a "picture" (raster) exercise.

The point was- there is no "math" to pull out- its possible to extropolate a range but not true precise dimensions. And they are estimates.

>>>If we're talking about rectifying the images, then IMO a DEM (digital elevation model) would be most appropriate, except we'd need a surface model of Patty rather than an elevation model. But this puts the cart before the horse since it's the model we're trying to determine. However, it's still more feasible than somehow deriving isometric transforms to do this.

Thats what I was saying.

As far as the overall 3-5%- that wasnt a mathmatical figure ( as I said earlier I never thought to do one and havent done one)- thats just an estimate ( off the hoof) based on doing it on parts with known dimensions when doing various back engineering designs on CNC's. It was not to be taken as the literal definitive.

Another problem with scaling ( which is based on the presumption it may be a suit as well as no known BF anatomy to base it on)- if its a suit- the construction and the fact its supposed to look like a BF builds in dimensions that when distorted may in fact appear to be out of human range when in reality it isnt

[Guest JohnWS]

There is a problem with the "yoke" of padding you referred to. Raising the shoulders can be done to apparently lengthen the arm (excellent example, the "Metaluna Mutant" creature from This Island Earth, a 50's si-fi classic) but there's a catch.

Here's the MM for interested parties who may not have seen it.

Nice bit of art & craft - budget seems to have run out at around waist height .

[Bill]

John:

Thanks for the Metaluna Mutand photo. Didn't have one in my own archives to show.

I did actually build a recreation of this suit in the mid 70's which is why I'm familiar with it's shoulder padding, and why I know it looks terrible if you swing the arm any way from straight down to a raised position forward, back, or sideways.

Bill