Topic: Digital Camera Pixel Size (Read 2505 times)

mkgillman · « **on:** January 24, 2007, 01:36:52 am »

I am currently doing some research on using off the shelf digital cameras to create accurate, photorealistic three-dimensional models for mapping purposes. The one step that is very important to acheive the type of results needed for my purposes is to identify and correct any lens distortions for whichever lens is mounted the camera. There are various types of software for this, but there seems to be an interesting "feature" on the current calibration software that I'm using in that it adjusts the size of the pixel for the camera. I am mainly working with a Canon 350D and Canon 50/1.8 lens.

From what I have found on the internet, it appears that the Rebel sensor has pixel sites that are 6.4 micrometres square. However, when I check this against the pixel resolution for the camera that number is almost exactly the (pixel dimension)/(sensor dimension). I was wondering if the pixel sites are the precisely formed onto the chip that they are all uniformly 6.4 micormetres square or if there is some variaton that would account for the calibration program modifying pixel size.

Thanks for any help and I'm sorry for the rather dry nature of the question.

-matt

Jonathan Wienke · « **Reply #1 on:** January 24, 2007, 02:20:14 pm »

There are two measure measurements you may be confusing: pixel pitch (the distance between the center of one pixel and the center of the next pixel), and pixel size (the diameter of a single pixel). Pixel size is always smaller than pixel pitch, because there is always a certain amount of space between one pixel and its neighbors, used for row/column wiring, reset & buffer transistors, and other things. The measurement you care about is pixel pitch.

Both are as accurate and consistent as the chip manufacturers can make them; chips are fabricated with a series of masks that must align very precisely for the transistors and photodiodes and wiring to connect properly.

mkgillman · « **Reply #2 on:** January 24, 2007, 09:18:39 pm »

The error that I'm seeing is a very small number (0.4-0.

micrometres, but I'm not sure if that is larger then the manufacturing tolerances.

The overall difference in the resulting calculations is rather small, but I'm trying to use the results for control in a research project.

I appreciate the help.

Jonathan Wienke · « **Reply #3 on:** January 25, 2007, 02:05:20 am »

The errors of the greatest magnitude by far are going to be from improperly corrected lens distortion, not sensor pixel dimension/spacing variance. The manufacturing tolerance for pixel spacing would have to be smaller than the variance you're talking about, or the circuitry on the sensor wouldn't be connected properly.

mkgillman · « **Reply #4 on:** January 25, 2007, 11:03:50 pm »

That is largely what I had expected and it seems that there were a few variables that were not being acounted for in the calibration procedure that I figured out today. I've got the derived pixel size to come down to around 2% of the nominal value. Thanks for the help.

Author Topic: Digital Camera Pixel Size (Read 2505 times)