Enzymind Labs

    A good looking corpse eye would not be so gruesome. I was thinking of a largish intact animal, where you might not be able to tell it was dead. The squirrel jerky is dead, no question. I've got some squirrel jerkey on hand. A good disection (frog, pig) would be good, a kind of virtual tour of innards. Good fur/feathers is nice in stereo, great optical properties.

    The rock with 10s exposure of water is a common technique. Every time I see one, I want to make a "power of 10" video -- 1/1000 to 100 second exposures with different lightings/flash to roughly match background illumination. Many at each exposure to show variance differences, sort of a calming of the waters from short to long exposures (time scales).

Just curious as to what you see as the next two or three steps in the DIS project are. I'm really eager to see any kind of example of this -do you think you will have to write a large portion of the core DIS software for a demo image or is there a way to do some moc-ups or examples using matlab or some photoshop layering stuff?

I think the wings would be fantastic, but I'm just wondering if subjects with a little more structure or surface contour might test/ demonstrate DIS more effectively. But then again I've really only seen drosophila wings close up (i was screening mutants for whether they had 2 or 3 hairs on a part of their body -that experience weened me from any interest in fruit fly genetics) and my recollection was they were remarkable smooth at the macro level (stereo microscope) but there were definitely little hairs an intricacies that would be fun to explore -and when you say array i think you mean take a dragonfly, beetle, wasp etc and images them all together. Just got a little portable photomicroscopy set up on ebay we'll see how that works with the xt. this this and that

     Hey, this thing would be handy for macro DIS if it could handle about a 500g camera without too much backlash/shake. The controller and feedback is what would take some time to do right. Could easily automate large number of acquisitions. With this memory (camera and offload to computer) becomes the speed limitation. 

    There's one minor issue with moving the camera in z (for DOF compositing) that I'll have to experiment with: if you move the camera, the perspective will change. No easy way to control the lens focus with comparable accuracy. 

Sorry I've been off the radar. A bit of a mental vacation going on.

    The next DIS software step is to make some real improvements to Space, such that DIS can be accomodated. These include some structural changes: 1) accomodate different voxel sizes transparently (display images at their appropriate scale, not at one pixel/byte but one pixel/mm) 2) add panning capability and 3) when displaying sets of 2-D images, change the resampling routine to speed up refresh. This last problem came up when making the DIS demo, and will make future demos much more interactive and friendly. I still haven't finished 32-bit grayscale (almost there), so don't hold your breath on these new things.

    I will make a DIS data set or two, macros, but I'm going to stick with 1-D rotations about a central point, because I can display these fast and they are mechanically more simple to make. And there are interesting ones to make.

    Here are the general guidelines: MOST IMPORTANT: Make good quality aquisitions (good photos, lighting, focus, even across images, etc.). Divide the memory you have by the size of the image you will store, and this gives the number of images (I think you will always be memory limited, either hard disk or RAM). The spacing isn't too critical, the closer the better (without running out of memory!) You can always only present every other one if you don't have enough RAM.

    If you want a good spacing for stereo, this also isn't too critical. You might take a test pair and view it. But if you give me the parameters (focal length, near and far points in the image) I can give you a good idea of what a good baseline (camera spacing) would be. I wrote a pamphlet about this, before I had a computer so I only have it in type. I'll give you a copy some time, or find a web reference. Basically you want the difference in angular disparity between the near and far point (when viewing the pair) to be between 1 and 5 degrees, and mostly on the smaller end of that range. Again, if you oversample (take images at short camera spacings), this isn't an issue because you have a choice of short or longer baselines.