As Dave explained, like pin cushion effects and tilt perspective correction, cropping shifts the image centre and the photomatch software assumes the image undistorted to perform the mathmztics…
How would the software know where your image centre was if the image was arbitaritly cropped…
The software cannot predict where you have cropped the image and the raw image centre is an essential factor in perspective calculations… try matching a one point perspective with the image cropped https://www.math.utah.edu/~treiberg/Perspect/Perspect.htm
I know that Match Photo works bestwith uncropped photos. I accept that, but I’m still with c.j.ryan in asking why?
Implicit in Dave R’s explanation is the notion that the edge closest to the camera must be in the center of the image, so that the vanishing points are equidistant from that edge. But it’s easy to take a photo with the close edge off-center, so the vanishing points will not be equidistant from that edge. Conventional SketchUp wisdom says this photo should work in Match Photo as long as it’s not cropped. But I wonder why it won’t work in Match Photo even if it’s cropped.
So, to paraphrase c.j.–Why?
Yes, I was going to chime in at the beginning, but I had a scheduled Zoom meeting to go to. I have lots to say, but it’ll take a little while for a complete explanation. I’m supposed to be prepping a presentation on all this for 3D Basecamp, and now who knows what’s up with that, but I’m still determined to find a platform to deliver at least some of it either way.
The quick answer is this; I believe the engineers who created the Match Photo tool simplified the problem for themselves by making some assumptions and reducing the problem at hand to certain special cases, namely two and three point perspective, but not one point perspectives. They also assumed that the photograph is a natural, un-doctored photograph obeying some simple rules of optics. I’ll bet you can crop a picture so long as the center of the cropped version is the same as center of the original image, because that would be the same as using a different lens or camera. 4x5 and other view cameras can do “perspective correction” through swings and tilts, which will mess with the assumptions behind the Match Photo engine, and that’s what creates a problem. There are PC lenses to do the same with regular, interchangeable lens cameras, and you can also achieve the same effect in software. If you didn’t take the picture, you can never be sure what was done with it.
I know next to nothing anything about photography or lenses (or Match Photo), but I can perceive the effects and understand the issue. It seems RTCool’s answer was well-implied many posts above-- Keep the image centered so the position of the camera is known and its effects can be predicted in the algorithms. Seems clear to me.