[opendtv] Re: Mediaflo Technology - Mark_Schubins_Contrast_Sensitivity_Function_Presentation

  • From: Mark Schubin <tvmark@xxxxxxxxxxxxx>
  • To: opendtv@xxxxxxxxxxxxx
  • Date: Sat, 11 Nov 2006 22:46:24 -0500

Tom Barry wrote:
For anybody still following this thread, Mark's Powerpoint presentation can now be found on my web page at:

Here's a description of the slides:

1 - This is a contrast-resolution grating as it is normally presented, with resolution increasing to the right and contrast decreasing to the top. You will probably see some sort of downward-shaped curve at the top of the grating. There is no such curve in the image. The curve is being introduced by your visual system. The exact shape and position of the curve will depend on your display, your distance from it, and, of course, your visual system.

2 - This simply inverts the grating so I cam label it properly. Resolution still increases to the right, but now contrast increases to the top instead of to the bottom.

3 - I have drawn in a contrast-sensitivity function. Whether it matches what you see will, again, depend on your display, your distance from it, and your visual system. But, wherever the curve lies, everything under it will be invisible to an observer with that contrast-sensitivity function.

4 - This is a typical modulation-transfer-function curve shape, but how broad the shoulder is will depend on the listed factors. The psychological sensation of "sharpness" is proportional to the square of the area under the curve.

5 - These are real-world MTF curves, as measured by Bob Atkins, who gave permission to include these. They are for the two Canon still cameras shown, with the same lens. There is a 14% linear difference in the number of photosensors in the two cameras. Even that small increase has made a significant increase in the sharpness. I'll send Tom another slide that shows that sharpness increase.

New 6 (not there yet, until I send it to Tom) - A visual representation of the sharpness difference of the two Canon cameras, based on a portion of an image of the Canon brochure,

Old 6 (new 7, when I send Tom the sharpness illustration) - This is the scary one. I've superimposed the hypothetical MTF curve over the contrast-resolution grating and the CSF curve. The two particular curve shapes and locations are hypothetical, but the fact that the wedge to the left is all that is available and visible is not.

New 8 - I've also added this slide with two composite photographs from Aude Oliva at MIT and Philippe Schyns at the University of Glasgow. Again, they are used with permission. You may read a paper about those images here:
You will probably see an angry man on the left and a neutral woman on the right in the top, larger images. In the lower ones, you will probably see a neutral woman on the left and an angry man on the right. There is no difference between the upper and lower images other than scaling. You may convince yourself of this by looking at only the top images and walking backwards from your display until they swap. There is a low-spatial-frequency image of a neutral woman on the left and an angry man on the right combined with a higher-spatial-frequency image of an angry man on the left and a neutral woman on the right. As you walk back from your display, the peak of your contrast-sensitivity function moves from the higher-resolution angry man on the left to the lower-resolution neutral woman on the left. The low-spatial-frequency information is eight cycles per facial image; the high is 24 (equivalent to 48 pixels across). The source images are 256 x 256 at eight bits.

Please do not re-use any of the images without permission. I will be happy to provide contact information to you.


You can UNSUBSCRIBE from the OpenDTV list in two ways:

- Using the UNSUBSCRIBE command in your user configuration settings at FreeLists.org
- By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word 
unsubscribe in the subject line.

Other related posts: