A Deconstruction of Eric Salter's Chopper 5 Rebuttal

by Ace Baker

8.01.07

Eric Salter has offered what he calls a rebuttal to my Chopper 5 velocity analysis.

"Rebuttal of Ace Baker's 'Chopper 5 Composite' Analysis" by Eric Salter

His entire critique is premised around a demonstrably false claim. Namely, Salter claims that the camera motion of the Chopper 5 shot is perfectly stable, and so would have a perfectly consistent effect on the apparent motion of the airplane. I invite everyone to please download the Chopper 5 video and step through it frame by frame, paying attention to the apparent motion of the twin towers, to see whether this apparent motion is perfectly consistent from frame to frame, or not. I am confident that everyone in the world except Eric Salter will agree that the motion is quite unsteady and inconsistent.

download Chopper 5 video clip

In addition to the false claim about the stability of the helicopter shot, Salter wrongly suggests that fine detail is more important than overall position in determining velocity, he dismisses the validity of my control cases based on an irrelevancy, fabricates a huge margin of error with no control case whatsoever, concocts a strawman argument and attributes it to me, incorporates the strawman into a false dichotomy, directs attention to an unrelated video claiming it would have had to have been made the same way as Chopper 5, all while concealing his stabilization data from us, so that we cannot actually test his basic claims.

Below is Salter's piece in its entirety, in shaded boxes, with my comments interspersed.

"Rebuttal of Ace Baker's "Chopper 5 Composite" Analysis

Many times in the past several years we've heard dramatic announcements of definitive proof of no-plane and video fakery claims. These previous false-alarms have proved to be mistakes and Ace Baker's analysis of the Fox 5 footage is no exception to this pattern.

Baker claims that the lower variability in the measurements of the plane's speed in his control cases prove that the motion of the plane in the Fox 5 footage is unnatural. This claim is in error. The control cases are not equivalent to the Fox 5 footage and do not shed light on the expected margin of error in the Fox 5 analysis. In fact, his control footage is sharper and lacks the noise and distortion (visible in the wavy, rippled edge of the WTC towers) of the VHS source for the Fox footage. The most important factor in stabilizing or tracking motion is the clarity of detail of the tracked object and the consistency of that detail from frame to frame and not so much on the number of pixels that object occupies. It's not surprising that he achieved a lower error rate with his control footage analysis.

That's false on two counts. First, my control case has considerably less detail than Chopper 5. My airplane was farther away (about 5 miles, compared to about 4 miles for Chopper 5), was shot with an inferior lens, and may have been a smaller plane to begin with.

Control case (L), Chopper 5 plane image (R), both enlarged 400%.

The difference in detail between the edge of my building and the edge of the twin towers is perfectly irrelevant. My building was much closer than the twin towers, yes, but we are studying the motion of the airplanes, not the buildings. The fact that Salter immediately directs our attention to an irrelevant detail in order to support his point betrays the nature of the disingenuous diatribe which is to follow.

Secondly, the most important factor in tracking motion is not clarity of detail, as Salter claims, but rather pinpointing overall position. The way to pinpoint overall position is to find the geometric center of the shape. This has to do with the number of pixels the object occupies, not necessarily the amount of fine detail. Salter has it backwards.

The increase in variability in Baker's post-stabilization measurements of the Fox 5 footage is not suprising either: stabilization of that footage would not change the variations in speed, except for some jitter caused by the software stabilization process, which would generate errors due to the low quality of the footage.

I stabilized mine all by hand, I simply compared every frame to frame 1, and got it as close as I possibly could, keeping a record of how far each frame was moved. At least I used the same method on each frame. Salter apparently used software stabilization, but then manually adjusted certain frames that he deemed to have "errors" in them. This sort of selective tampering is not necessarily suspicious on its face, but we must note that Salter has provided no record of how far each frame was moved, either by software or by hand.

Low quality footage? Let's not forget that Eric Salter is the one who named it "hi resolution excerpt". By the way, I'd like to formally request that Eric Salter digitize the entire Chopper 5 sequence in the same (or better) resolution as the "hi resolution excerpt". A lot of us would like to look very hard and see if we can see an airplane anywhere in the shot, apart from the one that jumps in at the last second, right after the miracle zoom.

The helicopter already had a mechanically stabilized camera system to remove shakes and vibration, leaving only a steady drift to the left, probably caused by the helicopter's movement.

Nonsense. The camera movement in Chopper 5 looks nothing like "a steady drift to the left". I invite anyone and everyone to please download Chopper 5, step through the footage frame by frame, and watch the towers. They will move considerably more between some frames than between others. It begins with a fairly steady drift from frames 1 - 4. Between 5 and 6 there is almost no change in position. Then the drift starts back up again, but gradually, eventually coming back to around its former rate. Between 15-18 it slows down considerably, then jerks. Steady drift to the left? Who are you kidding?

The goal of the on-board mechanical stabilizer may be to remove all the shakes and vibration, but that certainly does not mean it is perfect. Furthermore, even if we could perfectly isolate the camera from the engine vibrations, the entire helicopter is moving. In order to hover, a helicopter's rotors must continually "beat the air into submission". Helicopters cannot hold perfectly still. There are bound to be changes in the viewing angle, despite efforts to remove them all. Those changes are most evident in the Chopper 5 background, but they are not evident on the plane.

Stabilizing this footage removes the same amount of movement from each frame, leaving the before and after measured motion variations the same, except for the positioning jitter, which can be seen in the higher variability in Baker's stabilized velocity graph.

Now it becomes obvious why Salter is trying to get us to beleive that the drift to the left was steady, when clearly it is not steady. It is because he knows that variations in the rate of drift must cause variations in the apparent speed of the airplane. He wants us to accept on faith that "stabilizing this footage removes the same amount of movement from each frame" so that we will not wish to compare stabilized vs. non-stabilized versions.

So the only question remaining is: do these variations exceed what would be the expected margin of error in measurement of the Fox 5 footage? For this an attempt to calculate that error must be made.

The subjective placement of the wireframe over the plane image is definitely going to create some error. The following diagram shows 3 possible placements (at 800% magnification) of the wireframe over the image, one in a center position and two other positions moved to the left and right .5 pixel. The size of the pixels can be easily seen.

The center and left positioning are virtually indistinguishable, and the right adjustment only barely looks out of alignment. So the exercise of visual placement alone introduces a minimum margin of error of just less than 1 pixel.

And there are more factors which would add to the margin of error:

•Distortion of the shape of the profile of the plane from frame to frame due to poor quality vhs footage.

Not really. VHS does reduce horizontal resolution in comparison to broadcast standard. While this might tend to alter the shape of the nose, it does not tend to alter the overall position.

•Stabilization usually calculates stabilization at a sub-pixel resolution. I've seen jumps of a large fraction of a pixel when stabilizing high quality footage, yet the Fox 5 footage is noisy, poor quality VHS footage (in fact, I had to make adjustments to the stabilized Fox 5 footage by hand because of some errors in the stabilizing process, and that manual process also certainly has a margin of error). I don't know exactly what this additional error would be–only someone with more experience in this subject can know this value with some confidence–but a good sized fraction of a pixel seems like a bare minimum conservative estimate.

This fails to explain how I get so much less error from both my control cases. My control cases have less resolution (on the plane), are of smaller planes, and (in the camera shake version) was stabilized by hand. All of the same accuracy problems Salter mentions would be present on these cases, and to a larger degree if anything. In trying to explain away this contradiction, Salter asks us to believe that my control cases are of better quality than Chopper 5. As mentioned above, his basis for this claim is the difference in detail in the buildings, which are irrelevant to the motion of the planes.

Given the error in visual placement plus these other two factors, the margin of error should be at least 1 pixel and could possibly be higher. A velocity measurement uses two position measurements, therefore the minimum margin of error in the velocity measurement would be 2 pixels or higher.

Where is your control case, Mr. Salter? Having already used disingenuous means to attempt to discredit my control cases, Salter now asks us to believe a different, higher margin of error on his word alone. If legitimate videos have errors as large as you say, let's see them.

The following shows the stabilized Fox 5 footage with a wireframe overlay placed over the plane and a timeline showing the change in position and speed of that overlay:

http://www.questionsquestions.net/WTC/fox5velocity.html

Yes, but it does not show the per frame stabilization data, thus we cannot compare stabilized to unstabilized. Salter has missed the entire purpose of the exercise. He asks us to accept, again on faith, that the motion of the camera was perfectly smooth. Anyone with eyes can see that it is not, and I again ask everyone to step through the Chopper 5 video clip frame by frame, and take notice of how much the towers move. If we are to believe Salter, we must believe that the towers appear to move exactly the same amount between every pair of frames. Nothing could be further from the truth.

Frames 12 and 13 were excluded from the velocity calculations because the plane behind the graphic. The data for frames 1, 11, 14 and 22 might be less reliable than the other measurement because only a small portion of the plane was visible, but I included them anyway.

Now, if the plane is real and moving smoothly, the difference between two velocity measurements should not exceed 2 pixels. This is exactly what is the case, as shown in this graph of the velocity of the wireframe:

The largest change in speed between any consecutive frames is between frames 2 and 3, 6 and 7, and 18 and 19: a 1.6 pixels/frame change, which is less than the 2 pixels/frame minimum margin of error (corresponding to a .8 pixel error in position, which could be accounted for solely by the subjective error in positioning the wireframe overlay). Thus, the motion of the plane in the Fox 5 footage clearly matches the motion expected of a real plane.

What we see on Salter's graph are data far too unstable to come from a real airplane on stabilized video of this quality. At this same level of magnification, my controls would show deviations of less than 1/2 pixel.

Where are Mr. Salter's control cases? Where are his stabilization data? Where is his comparison between stabilized and unstabilized velocities? Nowhere that I can see.

Baker's measurements of larger errors are simply not representative of the Fox5 footage, and are a result of less accurate placement and stabilization (by the way, Baker apparently doubled the image size of the 640x480 footage before measurement, resulting in motion of 20-25 pixels per second instead of around 11-12. His values must be multiplied by .56 to be equivalent to the values in my graph above, which come from the 720x480 footage at normal scale).

Though doubling the size of the image does not add any meaningful data to the image, it does allow for more precise positioning of frames during the stabilization process, and of outlines during the position measuring process. Unlike After Effects, (which Salter used), Photoshop and Image Ready (which I used) do not allow sub-pixel adjustments. Theoretically After Effects could allow for more precise stabilization, and more accurate results overall. But we do not know what that additional accuracy might reveal in this case, because Salter has hidden from us a most important set of data. Namely, Salter has not shown us how much each frame was moved, so we cannot compare "raw" to "stabilized". We are simply left to take Salter's word that the camera motion was steady, when it is quite obvious that it was not steady.

One might argue that Baker couldn't have come up with such similar graphs of velocity for the stabilized and non-stabilized versions of the Fox footage unless those variations were authentic, but this would not be necessarily true: he could have simply made the same subjective errors in positioning the wireframe both times, resulting in similar graphs.

The positioning of my wireframes is perfectly consistent between the stabilized and non-stabilized versions, because I only positioned the wireframes once, on the stabilized version. My non-stabilized data was derived by starting with the stabilized data, and subtracting the distance each frame had been moved from the previous frame in order to stabilize it. If a particular frame had shown a stabilized velocity of, say 24, and that particular frame had been shifted 5 more pixels than the previous frame in order to stabilize it, then this would give an unstabilized value of 19. This crucial comparison is impossible to carry out under Salter's method, because he has hidden his stabilization data from us.

The fact that Salter simply assumes that there is no camera motion to deal with, and does not give his stabilization data so that we may check for ourselves, is the height of a thoroughly unscientific approach. Let me here formally request that Mr. Salter please send me a copy of his After Effects session.

In regard to the exit of the plane, the alleged "nose cone" emerging from the building is not the shape of the nose of the plane: it is very clearly thinner than the nose of the airplane when it first exits the building, then grows in size to approximate the shape of the nose, then continues to widen and starts to move down a bit, which is exactly the behavior one would expect from a smaller piece (or pieces) of debris pulling an expanding cloud of dust or smoke behind it. It simply can't be part of a keyed plane image or CGI model. Because the shape is changing Baker's motion tracking data of its velocity is pretty much meaningless, and furthermore that data has generally the same amount of variability as Baker's measurements of the plane, which means it's probably not indicative of any abnormal motion for the same reasons the plane data is not.

Salter has seen fit to stop his study before Pinocchio's nose emerges. So we are left with nothing whatsoever to go on other than his say so. Whether the phenomenon is actually the nose of the airplane, or smoking debris, it displays behavior obviously unphysical in nature. It springs out from behind the building (not the side of it), it nearly stops moving, it speeds up again, and it goes up and down just for show.

Certainly there can be subjectivity in marking the overall position of this moving object. But I did so in order to quantify what is already completely obvious to see. I implore everyone to please look at Pinocchio's nose, and observe the motion. Does it appear to move smoothly, or does it jump around?

What is the likelihood that a random smoking chunk of debris would so closely match the size, shape, color, and altitude of the airplane nose? When we look at the other debris shooting out of the towers, does any of it match the size, and/or shape and/or color/ and/or altitude of the airplane nose? Nowhere that I have seen.

And why would this debris have the same sorts of motion problems found on the airplane, when the rest of the debris does not have these problems? If would be nice if Salter had provided his Affter Effects study on this, but alas, he has chosen to truncate his timeline before this event occurs. I can't help but be reminded of NIST truncating their timeline as soon as the twin towers begin disintegrating. How convenient.

So in summary:

•Baker's control cases are not equivalent to the Fox 5 footage and don't contribute anything to the calculation of a margin of error.

At least Salter is not disputing the low error rate in my controls. My control cases are of worse quality, and should display higher margin of error than Chopper 5. Salter only claims my controls are of better quality than Chopper 5 by referring to an irrelevant feature, the closeness of my building, rather than what is important, the distance of the plane.

•A conservative estimate of the margin of error in positioning the overlay on the Fox 5 footage must be around 1 pixel (or more).

Salter can employ whatever adjectives his wishes in describing his method of estimating his error. In the end, he has supported his position with no control cases of any sort. I support my different, much lower error findings with two control studies.

•The motion of the plane in the footage is well within this margin of error, therefore the motion is natural and there is no evidence for video fakery.

Salter has shown nothing at all. I repeat my requests for Salter to publish or make available his After Effects session, in particular his per frame stabilization data, which will tell us how much each frame was moved in x, y zoom, and rotation. I also repeat my request that Salter digitize and make available in hi resolution the entire Chopper 5 shot, for our continued analysis.

There's really no need to read further but I have some more observations for those who want to explore more. In addition to technical mistakes Baker's article is filled with biased and illogical thinking. His arguments show that he doesn't understand how keying and compositing is achieved on a professional level.

Translation: Ace Baker doesn't know the serial number on the gizmo (apology to Judy Wood for borrowing the phrase).

His idea that footage of a real plane was chroma keyed live is absolutely silly.

Strawman. I never said the plane was chroma keyed. I clearly state in my paper, that the plane image would be prepared ahead of time, and placed by itself on a transparency. This is called an "alpha channel", as Salter is about to point out. I did not metnion the term "alpha channel", as I didn't think it necessary. More important is to describe the principle involved. I'll make sure future updates of the paper include the term.

It would never be used when photo-realistic CGI would better accomplish the task (unless the perps deliberately wanted to emulate 1950's quality special effects and expose themselves).

Proceeding with his chroma key strawman, Salter presents a false dichotomy: Either the plane was a chroma keyed live plane video, or else it was a CGI on an alpha channel. Missing in Salter's deceptive dichotomy is the possibility that it was originally a real plane on video, which had been painstakingly removed from its background, and placed onto an alpha channel waiting to be inserted. This is the method I describe in my study, yet Salter is busy running amuck with his fabricated strawman and his false dichotomy.

Not only could CGI be perfectly synced to the motion of the background footage, . . .

Anything placed on an alpha channel can be synced to the background footage as well as anything else. The original source of the pixels residing now on the alpha channel do not matter. What Salter means by "perfectly synced to the motion of the background footage" is not clear. If he means that exactly one frame from the alpha channel will be overlaid onto exactly on frame of the background, and that these composited frames will be played at the excat right frame rate, then yes, of course the overlay could be perfectly synced.

But it seems Salter is trying to have us continue believing the lie he began his whole paper with, and that is the notion that the helicopter background shot is perfectly stable. It is not stable, and the motion of the camera during this shot could not have been known in advance. In that sense, it is impossible for the overlay to be "perfectly synced to the motion of the background footage". Rather, the two video sources will operate independently of one another - the motion of the plane is what it is, and the motion of the background is what it is.

it could be cleanly overlaid, eliminating the problems with edge details that go along with chroma keying.

Mr. Salter should re-read my paper. I clearly describe preparing the plane video transparency layer ahead of time. The "edge details" can be massaged with innumerable tools, for as long as necessary. The overlay can be tested ahead of time in "battle situation" by attempting to overlay it on a test shot of the twin towers. I do not speak of chroma keying the plane. I speak of chroma key, luminance key, find edges, etc, as a method to create the real-time mask of the towers.

Keyed real footage would only be used when both pieces of footage are shot using the same motion control camera movement, which would match the motion of each perfectly and eliminate motion anomalies of the type he says can be observed (and which aren't there anyway).

Computerized motion control camera booms and tracks exist that can learn and repeat complicated camera movements. These are mounted on terra firma. It's not possible to match motion on an inherently unstable platform like a helicopter. If you had a such a motion control system on board a helicopter, and repeated a camera movement, the two shots would have identical camera movement, except for the motion introduced by the helicopter itself.

Besides, if the anomalies aren't there, why do we measure them? Salter says it's all measurement error, but offers nothing to support this assertion.

And if they did use live footage, it would not be keyed live but be prepared ahead of time with keying and/or rotoscoping to create an image digitally formatted with an alpha channel for a clean overlay (and they could have easily fixed any variations in movement during that process).

Of course they could have fixed any variations in movement of the plane overlay, but that would have been a dead giveaway. A perfectly stable airplane path coming from an inherently unstable helicopter platform is an impossibility.

Besides, in broadcast video, the "Sportsvision" technology he cites in his article is used with CGI images, not chroma keyed footage.

In principle, the "Sportvision" technology could be used on any overlaid video source. However, the Sportvision "1st and Ten" system could not be effective on a helicopter. The system contains sensors in the base of the camera assembly that detect any panning, tilting or zoom, and feeds that information to the computer. The software can then move the virtual 1st down stripe correspondingly. The base of this system is on terra firma. If mounted in a helicopter, while such a system would still detect panning, tilting and zooming of the camera, it would not detect movement of the helicopter itself.

The pink-colored artifacts he mislabels "digital paint" are artifacts resulting from noise, poor quality VHS recording or compression.

Well, which is it? Noise? VHS? or Compression? I've yet to be able to duplicate the effect with compression and I have tried. Compression tends to reduce the number of colors needed, and to make similarly colored pixels the same. It does not introduce new colors, like pink, into areas where there was no pink before. VHS reduces horizontal resolution, and makes up the width by increasing the horizontal size of dots. This does not seem like a good candidate to introduce pink coloration behind the airplane wing. So we're left with "noise". What sort of noise Salter might mean, or how that would produce pink coloration, selectively behind the airplane wings, is not known.

Like all the other no-plane arguments, there are fatal logical flaws to the scenario Baker argues. If an overlay was used, the Fox 5 footage shows that the overlay was perfectly synced with the WTC towers: jets of dust come out of building exactly where and when the plane hits (propelled by last second operation of jet engines, as seen in the Fairbanks footage).

First, Chopper 5 does not show a plane hitting a tower. Period. It shows a plane slipping behind the edge of a tower, just like all of the live plane videos. Thus it did not require perfect synchronization with the explosions, it just had to be fairly close. This is why Salter chooses here to refer us to a completely different video, "Fairbanks".

Evan Fairbanks, of course, infamously said on national TV that his footage and the plane crash looked like "a bad special effect". The Fairbanks video was not shown live, and has no necessary relationship to Chopper 5 at all, in terms of methodology. They had plenty of time to synchronize Fairbanks, Ghostplane, and any other videos which show the plane disappearing through the wall like a bad special effect. The fact that Salter must yet again divert away to unrelated, irrelevant items to support his point is illustrative of the utterly vaccuous nature of his position.

This tells us that the overlay of the plane would have been placed exactly where the alleged planted explosives would have been in the building, confirmed again by the flames coming out directly opposite on the other side of the building a split second later.

Whether it was pre-planted explosives, an incoming missile, directed energy weapons, or some combination, there was enough time to record the actual live event, rewind the file, position the overlay both in both space and time, then hit play and broadcast the composite. The is a scenario suggested by SocialService in "September Clues". The video documents what appears to be a 17 second delay between the live event and when it was broadcast.

So the timing and positioning of overlay would have been precisely coordinated with the camera movement.

How? How could the overlay be "precisely coordinated with the camera movement" on a drifting helicopter? I don't know. Salter does not say either. My data show that the timing and positioning of the overlay most certainly were not coordinated with the camera movement.

Therefore (since they would have used CGI and nothing else) it doesn't make sense that there would be any irregular movement of the overlay as Baker suggests.

I didn't suggest there is irregular movement, I measured and documented the irregular movement. So does Salter, please see his graph. It makes perfect sense that there would be irregular movement to the overlay, because, as mentioned, removing all camera movement from the overlay would have exposed the plot for certain. There is no way to ensure a perfectly steady shot from a helicopter. A perfectly stable plane motion coming from a helicopter shot is impossible, and the perpetrators knew this. Salter knows it too.

And it also follows that it doesn't make any sense at all to suggest that the conspirators would mess up the exit of the model from the building when the entry was absolutely perfect, AND furthermore to suggest that they remembered to slow down the overlay during it's penetration of the building (as seen in Baker's velocity graphs) yet forgot to have it actually stop within the building. It seems that no-planers simply can't think logically.

The video does not show "entry", perfect or otherwise. This is a repetition of a psychological tactic attempting to convince everyone that Chopper 5 shows a plane hitting a tower. It does not, and neither do any of the "live" plane videos.

When preparing the alpha channel overlay, excess frames are required. They wouldn't know precisely which frame was going to be the last to disappear behind the masked edge of the tower. I believe the original footage depicted a plane flying right to left and banking to the left. As the shot progressed, the camera began tracking the plane, thus causing it to appear to slow down, and bob up and down slightly.

During the live overlay, the perpetrators had a window of opportunity to stop the motion of the plane. Stop it too soon, and the world sees a frozen or disappearing plane. Stop it a little too late, and you see the nose peaking out the back side of the mask, just before the explosion starts on that side of the building. The latter is what happened, of course. Rather than stop or disappear the plane, they did a quick fade to black. Notice that Salter does not comment on the fade to black.

Baker suggests that clear weather was essential to the conspirators plan (because of their rather unprofessional plan to use chroma key). And just how did they ensure clear weather for a date planned months in advance? Was God in on the plot? Amazingly, this is even more ridiculous than the previous no-planer assumption that the conspirators could have controlled all the cameras in New York on 9/11.

The plane was on an alpha channel transparency, as mentioned. It is the edges of the twin towers which must be identified in real time, using chroma, luminance, etc. A dark, gray, foggy day, for instance, would make identification of the dark, gray tower edges difficult. The perpetrators may very well have been willing to postpone 9/11 had the weather not cooperated. I am not aware of any "previous no-planer assumption that the conspirators could have controlled all the cameras in New York on 9/11". This sounds like another Eric Salter strawman, no surprise.

In addition to his own claims, Baker cites a series of no-plane arguments (such as that the plane in the Naudet footage is smaller than a 767) which not only have already been thoroughly debunked (and a long time ago too) but which were debunked with analyses much simpler than the kind he himself attempts in his article. Because of this, one is forced to assume he possesses not only a pathological level of intellectual dishonesty, but also a zealous bias towards the no-plane scenario totally in contradiction to the scholarly tone of his article.

Once again, I felt oblidged to revisit this issue and conduct this analysis, since this analysis is outside the expertise of other 9/11 researchers I know who are working in this area. Unfortunately it was a royal waste of time, delivering totally predictable results. The lack of a logical scenario for the no-plane theory should be enough to invalidate it, but many people don't think logically and, sadly, propaganda like this still has to be rebutted, as the unwary will be fooled by it's veneer of authenticity.

So, having based his entire paper on the false premise that the helicopter shot was perfectly stable, fabricated a huge margin of error with no control cases, concealed his stabilization data, employed the strawman argument of chroma key, offered a false choice between CGI and his own strawman, diverted attention to irrelevant details and even to a completely different video; Salter then has the audacity to accuse me of intellectual dishonesty.

Let me close by repeating my request that Mr. Salter release a high quality copy of the entire Chopper 5 video, and any other 9/11 videos he may have in his possession. I'd like to see if I can find a real plane anywhere.