chasing unreality

[liveart]

shifting the viewers perspective... isn't that what art is all about? Here is a bit of fun we can have with SA's 2d blur function. It relies on a illusion known as tilt blur. start with a bird's-eye view of a real scene.. a parade, a view of main street, a view of the city's roof line.. anything you have with a slightly downward viewing angle. With this as your source image lasso select the center of focus in the image. Feather the selection twice. Invert the selection. Now head to the adjust panel and select 2d blur.. make sure your selection masking is on and very slightly blur the image in both axes. The illusion will work best if you blur just a bit more in the vertical than the horizontal. If you over blur you can either undo and try again or use the fixed image imop to blend back the original at about 40% release the mask. Voila .. instant miniatures... Now if you are Paul Perlow you can add to the illusion with hyper-real colors... ...experiment a bit with choice of focal area i.e. the area you mask out.. have fun...

[liveart]

[Synthetik]

The series of images you're been posting is pretty interesting. Do you have any links to the visual psychology behind this particular effect, or a short explanation of why your brain is fooled by these images?

[liveart]

John: Wiki entries on tilt-shift photography and the Scheimpflug principle will get you started. I have not read any papers on the psychology of why the brain is fooled, but certainly it has to do with field of focus.

pretty cool that SA can easily be set up to automate the process.. though as usual a little specific attention to each image does give better results.

[Synthetik]

I thought i'd throw out the idea of processing an appropriate movie file with this technique. Perhaps somebody reading can try that out and show us the results.

[liveart]

To do an animation the technique I would use would be slightly different. I would use spatial frequency mattes as an approximation of z-depth buffers and then use those to composite the blur. This isn't fully necessary but would more accurate in a moving image. Now if SA has real access to the z-depth of the image via a method I am not aware of that would be even better.

still I do think that the method I describe above could be used fairly well. I did make a PASeq, as can anyone who history records the described steps. The PASeq can be used to process a video.

The effect is not as strong if the camera plane and the center of focus are perpendicular.. that is if the camera is shooting directly at the object of interest. If anyone has seen the series "dollhouse" the opening credits are done via this sort of technique.

[Synthetik]

Apparently it's also more effective if you don't blur part of an object, so ideally you'd want to matte any objects at the boundary of the blur so they are either blurred or not.

I do think that what you are doing above might work fine for processing a video without getting into z buffer techniques. Like if you matted out the people in the center of the photo that are in focus, and then put them on top of a blurred picture of a chair in a room, and then used video for the in focus center part. If that makes sense.

[liveart]

it gets complicated when the z plane has an object that runs the length of the x axis. For example a flag pole. Ideally the pole would be either entirely out of focus or in focus depending depth of field's relation the the pole's z plane. But in the quick method I described .. either one would have to hand matte each frame or accept that the pole might be out of focus at the bottom and top but in focus at the middle.. not quite right.

Similarly the admonishment regarding object being partially blurred applies more to the z plane as normally the shallow depth of field would blur objects that were away from the center of focus even if they are in the same x-axis. So an object like a super-tanker or a building might very well be both in and out of focus if it were to take up enough of the frame.

another point is that SOMETHING has to be in focus or you just have a blurry picture. With a still this is no problem of course but in a movie with a fixed matte it might be very possible that nothing of interest is in the matted area.

So using a z buffer approach is the cleanest way .. though tons more work.

One more thing in a movie time lapse or simulation of time lapse helps the illusion

[Florence Artur]

I'm sorry, I don't get it. What is this supposed to do?

[mark g]

heres a tilt shift film


http://laughingsquid.com/keith-loutit-makes-a-beautiful-tilt-shift-...

[liveart]

Mark, great link and a perfect example of what I was saying about z plane issues. Re-watch the video paying attention to the light-poles, you will see them in focus in the middle and out of focus at top and bottom...

I will post a preset that will be a good starting point. Anyone wanting to get superior results will have to Adjust the masked area when the sequence calls for it.