In camera calibration (camera resectioning process, e.g. camera calibration with OpenCV) does the result for focal lengths fx and fy depend on the photo-sensitive area inside the pixel ?

If I have a camera sensor (CCD/CMOS) with a perfect square grid arrangement of pixel elements, but each pixel element has rectangular pixel sensitive area (photodiode). For example, pixel dimensions are a x a, but photosensitive area dimensions are 0.5*a x a. Would result of camera calibration still be fx = fy ?

And what if the pixel sensitive area has some arbitrary shape due to additional electronics in CMOS ?


The resectioning pixel grid is relative to the pixels in the output image being evaluated. It is not inherently dependent on the image sensor's pixels/photodiodes per se.

I.e. you can have pixels with .5a sensitive area (e.g. masked phase detection photosites) and the other portion of that photosite is calculated/mathmatically replaced in the output (or the entire pixel in some cases). Or you can have a pixel that is made up from two photodiodes of .5a that are binned together in the output (e.g. Canon's dual pixel phase detection). And neither would affect the resectioning calculation.

A film image can be resectioned as well... you just need to scan it into a pixel format.

The pixels in an output digital image are always square, as are the pixels of the device displaying the image.


I'm not familiar with camera calibration, but in any case focal length, the real focal length, does not depend on shape or size of the sensor.

Field of view does depend on shape and size of the sensor. And often we use 35mm-equivalent focal length as a measure to compare the field of view of different setups. But that's not the real focal length.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.