A way to measure force...

[KenWittlief]

Originally Posted by sanddrag

I like how they can take a photo, zoom in on a person's eye, say "let me see if I can enhance this" and then seconds later they have a gigantic and perfectly clear picture of what the person was looking at (the reflection in their eye).

I think the first time this was done in a movie or TV show, it was a photograph. with a good quality 35mm camera this is possible

in fact, I have a photo of the steering wheel of a Coast Guard 44 footer. If you blow up the 2" diameter chrome center of the wheel you can see me sitting in the pilots chair, pointing the camera

but I know I have seen it 'used' in tv shoes with video from a survalince camera. Video is 640x480 at best, what you see is all that is there - you cant zoom in and see what was between the scan lines of the video camera (theres nothing there)

if you use a fish scale you are measuring the impulse - the response of the spring in the scale, so the max reading would mean something if you could capture it, but the reading on the scale itself would not be the maximum force applied, because the scale is not intended to be used on moving objects

you would have to experiment with the scale to be able to interprete what the readings mean

but as a relative measurement, you could say impact A was greater than impact B. How much greater would take some work to figure out. I dont think the readings would be linear.

[Kevin Sevcik]

It's still energy as long as you know physically how far the scale's deflected as well. You'd be measuring the maximum potential energy stored in the spring. PE=0.5*k*x^2. You don't know k, but F=k*x so PE=0.5*f*x. Only problem is if ball bounces off in random directions, imparting an inconsistent amount of energy to the system.

You could also come up with a comparison is you don't know the deflection, using an arbitrary constant/guess for k and saying PE=0.5*F^2/k.

[dlavery]

Originally Posted by KenWittlief

but I know I have seen it 'used' in tv shoes with video from a survalince camera. Video is 640x480 at best, what you see is all that is there - you cant zoom in and see what was between the scan lines of the video camera (theres nothing there)

Not to hijack the thread, but that is incorrect. There is more "there" there than is there.

In simple terms, the pixel represents a region of data in the source image, and not just an infinitely small single point at the center of the pixel region. The information assigned to a pixel is based on the aggregation of the image properties across the entire image region assigned to the pixel. Using techniques such as superresolution processing, cascading image enhancement, multiframe quiver analysis, and multiscale morphological smoothers (typically nonlinear smoothing filters for detail enhancement at large scales), information can be extracted from a digital image "from between the pixels."

For example, imagine that there is a region that is composed of a white square surrounding an inner red square. The outer white square completely fills the region that comprises the pixel. If the center of the region mapped directly to the pixel, then the CCD pixel would just "see" a point at the center of the inner red square, and record "red." No other information would be recorded, or available. However, in the real world, as the light from the region passes through the optical system and impacts the pixel on the CCD it is aggregated together and the pixel records "pink." When this happens, some of that information about the source region is available to be extracted. This also holds true for a series of adjacent pixels. Using this property, it is possible to determine precisely where dissimilar regions fall between pixel coordinates. From that information, image details and geometry that is not readily visible in the original image can be reconstructed.

Image processing labs have been using advanced versions of these techniques for years to extract information from digital images. The techniques are commonly used at places like The Multimission Image Processing Laboratory at the NASA Jet Propulsion Lab, the previous Intelligent Mechanisms Group at the NASA Ames Research Center, and many of the intelligence community facilities. Details on the fundamental algorithms associated with digital image enhancement are found in virtually every set of proceedings of the annual SPIE Conference or any of the several photonics society newsletters.

-dave

[Dick Linn]

Check this: http://www.gigapxl.org/

Go to the image gallery, pick any picture, then scroll down and look at succedingly zoomed views and look at the detail. In this case, I don't think they are enhanced in any way. It's equivalent to a 4 Gigapixel camera! Now think of what detail image enhancement could show.

Back on topic, just shoot the balls into a pendulum of known dimensions & mass and calculate from there. Now think about what the mass of the pendulum should be relative to the projectile to achieve maximum transfer of momentum and avoid losses.

[KenWittlief]

Originally Posted by dlavery

In simple terms, the pixel represents a region of data in the source image, and not just an infinitely small single point at the center of the pixel region. The information assigned to a pixel is based on the aggregation of the image properties across the entire image region assigned to the pixel. Using techniques such as superresolution processing, cascading image enhancement, multiframe quiver analysis, and multiscale morphological smoothers (typically nonlinear smoothing filters for detail enhancement at large scales), information can be extracted from a digital image "from between the pixels."

I guess I should have been more specific describing the techniques 'used' on these type of TV programs. A surveillance camera will capture someones face in one or two frames, maybe filling in 25% of the screen (whole face is ~200 by 200 pixels)

they freeze on that one frame of video and zoom in on the center of the persons eyeball (which might be 5 by 5 pixels)

and then zoom in till that center black area fills the screen, hit the "image enhancement" button, and there you see someones face you can clearly recognize.

Not going to happen. The techniques Dave is talking about are common with big screen TVs, and projection video systems. When video is moving there is more data present, and you can interpolate pixels from frame to frame, and in 3x3 or 4x4 pixel blocks to determine what is between the pixels (as the image moves and pans) to scale the blocks up to 5x5 or 7x7 pixels. This is zooming in roughly 2x or 3x.

in fact, if a camera is panning sideways you can grab successive frames and get stereoscopic 3D images from a single camera (similar to the way an owl bobs its head from side to side to increase its depth perception)

OK, so yes, you can take multiple frames of video and pull more info out, and you can interpolate and project what is probably between the adjacent pixels

but you cant zoom in on video 100x and clearly see a persons face reflected in a 5x5 pixel eyeball

but with film, it is possible. A 35mm frame has the equivalent resolution of 10M pixels, compared to 0.3M pixels for a high quality NTSC video single frame.