frame alignment

[John Dolva]

Building on headshot head movements

I remember Craig posting something about the changes that happen when you shoot the same scene from the same place but with a tilt along any axis of the lens. This would be the same thing that is shown by the 'backyard photos, wildly different perspectives and size distortions from 'the real' by simply moving the lens.

So when you get a series of photos like a film, and if you align the frames like in a projector on to a static screen, unless, as is done when the camera has been held in a machine of some type that locks its movements into a totally jerk free lens, you get a film that jumps all over the place. The question then becomes how to smooth it out so it becomes more useful to answer what happened at a particular moment.

In the past (I talk of it as past, by which I mean pre may 16, 2006), this has been attempted by aligning the frames so they follow the contours of the street. That's ok for a very zoomed out view, however : still on a static screen.

If one had wanted to check out what happened in the limo, one would realise that the limo itself moves, so one might align the limo, thus changes in direction appear to be revealed by checking this alignment against what is happening to the kerb for example. Which is fine, except the lens has changed its plane, (or its attack).

How to compensate for this? By logic preferably.

Straight lines REMAIN straight exceopt that, depending on which area of the lens the photons pass through, they are PRESENTED as curves. This curvature is a product of lens structure.

Therefore, if on wishes to compare two lines like the kerb or a limo feature in close up, one must remember to deal with them as curved lines while remembering what they really are. And in nature nothing is really straight. As one zooms in closeup on a particular area one finds a need to start rotating the frames, which starts to create a huge jumble of movements in peripheral items. If one aligns the frames in this way and locate them frame by frame by comparing the most nearby background features, one gets a relative location for each frame. This then becomes a basis for the next step.

Edited May 20, 2006 by John Dolva

[John Dolva]

which is to choose an area to zoom in to and study. In this case a couple of lines that sent photons travelling to the lens which determined where on the film the photons would hit. Thus a film frame is a multivalue scan of the lens structure. A poor plastic lens would not compare to a glass one, a good glass in expensive amateur cameras are vastly different from that of pro cinematograph, which in turn is nothing comapred to a huge telescope. So here we have 500 odd scans of Zapruders camera lens. As far as being scans of the lens, and any discolorations and obstacles like dust and fingerprints, two frames are sufficient to 'see' the lens.

[John Dolva]

The next thing to keep in mind is that the film shows in the old alignment a jerkyness that is hard to explain. It helps to realise that this is not because the camera is being 'jiggled' around. In a macro sense perhaps, but here we are dealing with a mini time frame. Between frames,what produces the vastly differing poerspectives is better described as a swinging, sweeping (remember it's chaanges between frames in split seconds, gross human movements at this time for a camera person that keeps the subject in frame) movement. Once one gats the right alignment, one can even see the timegap between frames as a ribbon of totally different picture, a time when things were chaanging but no record was being taken of whatever was happening.

for the next step think zoom...(for some the penny may have dropped, but I found 'zoom'to be a major part of a leap from faith,incredulity etc to ok, lets talk.)

[John Dolva]

think zoom and 'correction'

stepping right back and looking at a series of frames from the Wiegman film.

next the curly swinging sweeping fading and enhancing (when correctly aligned, zoomzoomzoom)

subject of blur...

Edited May 21, 2006 by John Dolva

[John Dolva]

every minute item on the two succeeding frames has a blur associated with it. it's a slice of time, it has thickness. a beginning and an end.

as the camera lens swings (this concept needs developing as three or more types of swing are issue, the camera may tilt but track, the camera may be stationary and not track, it may be moving and not track)frame by frame it records blur while the film is being exposed.

if you take two successive frames the end of one is close to the beginning of the second.

so zoomzoomzoom, rotate, and bang, there's the sweet spot for that particular location. the blurs amplify what is unseen and the frames are aligned in that spot.

Because the relationship of the lens is continuously changing, once you have alignment to one particular zoom level, you can then zoom out, and see such things as the lens, its tilt and the timegap between frames. I think this is when one discovers the true interface curve where the images are the same.

[John Dolva]

...for that particular area, if you have taken great care, stepping outside the square with centre of rotation in paris(or timbuctoo for that matter) the heads are as they were seen by the two lens positions, to align them now in space and time look at the background immediately near the head, and locate sameness. ok

stepback this is the correct alignment for that particular interesction on the two lens tilts., the two heads are where they were in relation to each other 1/183 seconds apart. ok

how did they change between frames?

just pile them on top of each other remembering that you need to see the point of rotation way out in top field around which frame swings in over the other. finetune,repeat,study the results as transparent animation slowly going from frame to frame, adjust etc, analyse etc

http://educationforum.ipbhost.com/index.ph...indpost&p=62861

Edited May 21, 2006 by John Dolva

[John Dolva]

an ongoing attempt to conceptualise the thing is kind of represented by this, I wonder if it's possible to formularise such an effect, standardise, and perhaps even softwareirise .. eh? for auto zoom to any area?

Edited May 24, 2006 by John Dolva