Are pixels square or rectangular in their angular field of view?

HUDguy · Feb 4, 2013

I would like to try to make reliable size measurements of physical objects by imaging.

If I take a photo with a digital camera (like a "smartphone"), can I then count on that the pixels represent the same angle of view in horizontal as in vertical directions? For example, if I take a photo of a square straight on, perpendicular to its center, will it then measure as many pixels horizontally as vertically in the image which the camera produces?

I've made a test and the difference is less than 2%, which is within my margin of error for measuring the actual physical object and angles involved. I used 1600x1200 with a mobile phone and the "field of view" in terms of inches indeed was very close to that 1.333 proportion (up-down/left-right).

However, it strikes me that maybe this is an artefact created by the JPG-format I used to view the image, or maybe even earlier in the cameras internal automatical image processing. That the software tries to make it look better that way. Might I get different proportions if I use another image compression format?

And just for curiosity as a follow up:
If I would have the most raw form of image, say YUV or Bayer, would it still be true that the number of pixels per radian horizontally would be the same as the number of pixels per radion vertically? Are the optics and sensors design that way physcally? Or is it software managed afterwards?

Thanx y'all!

spacefuzz · Feb 4, 2013

Well each pixel size is dependant on the sensor used to capture the image. If you add jpeg compression to it, it will downsample based on its equation and I am not sure if it does so in a predictable way. I would certainly expect different results using different compression methods.

You can certainly use pixels as a way to measure things, and we used it often in the lab. To do so though we used a fixed camera at a known distance from the object we were deforming. Prior to running our experiments we would calibrate using a scale, and input that info into a program that counts pixels. The software would then calculate the total displacement over our tests. It was a free software, I want to say red was in the name.

amolitor · Feb 4, 2013

Pixels on the sensor are square.

Life gets weird and insane if they're not. There are probably scientific sensors where they care about resolution more on one axis than the other, but consumer photography ain't that.

To make reliable measurements you need to understand the magnification that your lens is giving you, and the pixel pitch on the sensor. These are both easily calculated, within a modest margin of error. The largest sources of error is likely to be the focus distance of the lens, and the actual distance to the object. If you have a fixed setup, you can calibrate those away -- shoot a picture of a ruler, then shoot the target object.

You can do quite accurate celestial navigation by photographing sunsets, interestingly, which is basically the same problem.

There are better ways to measure stuff, though.

snowbear · Feb 4, 2013

The shape of the pixels (square vs. rectangular) may (still) depend on the device displaying the image. At least on older monitors and printers, a circle on the monitor would print as an oval unless the aspect ratios were taken into account through software.

BrianV · Feb 4, 2013

There are camera and lens combinations that can be used for accurate measurements, but most are going to be approximations.

In addition to Interpolation between pixels, most lenses have geometric distortion. terms like "Barrel Distortion", "Pincushion Distortion", "curvature of field".

There are cameras made for use in the Lab, on microscopes, and such that allow precise measurement. A Monochrome Digital Lab camera with a Macro lens would be best suited for this job. They are not cheap. Best to use Digital Micrometers for accurate measurements. I'm old enough to have traced objects with an Analog all-mechanical Planimeter to get accurate measurements.

Electrical Planimeters,Planimeter,Placom,Lasico,Digital Planimeter - Tiger Supplies

I would have been happy with a modern one. BUT- these days, i would go with the macro lens and Lab camera.

KmH · Feb 4, 2013

You would need to check the particulars for which ever device you are using as far as pixel shape.

Some camera image sensors in the past used rectangular pixels.

A pixel is actually a photodetector and a transistor array. - http://en.wikipedia.org/wiki/Active_pixel_sensor

amolitor · Feb 4, 2013

I own two Lasicos. I think. Maybe I sold one, hmm, since I only see one on the shelf..

BrianV · Feb 4, 2013

Too funny- I think the one I used was from the 1930s.

The new Fuji sensors use Hexagonal pixels, and the Nikon D1x used rectangular pixels An old Kodak DCS420m (monochrome) with a micro-Nikkor would probably do a job like this, be fairly cheap, and the data sheets were published.