Re: polar to rectangular cordinates
Posted: 2008-07-31T11:10:27-07:00
You realize that this is only an approximation even for a linear (equidistant) fisheye lens. The resulting panoram is equally spaced (pixels) in angular units and does not conform to true perspective geometry. If it is other formats such as equalarea (equisolid), orthographic and stereographic, you will get other vertical distortions remaining.
A better approach is to do a proper fisheye to perspective view on a subsection of the fisheye image rather than rectifying the full image into a panorama and then cropping it to the desired subsection. I may get around to modifying my current defisheye script do to that as well as make proper panoramas for all of the above fisheye types.
In the meantime, Anthony and I are discussing implementation of the rect2polar and polar2rect scripts as true IM distorts. Hopefully we can add that soon. We are in the process of adding quite a few new distorts. Anthony can give you more details about what is currently under way and plans for the near term.
A better approach is to do a proper fisheye to perspective view on a subsection of the fisheye image rather than rectifying the full image into a panorama and then cropping it to the desired subsection. I may get around to modifying my current defisheye script do to that as well as make proper panoramas for all of the above fisheye types.
In the meantime, Anthony and I are discussing implementation of the rect2polar and polar2rect scripts as true IM distorts. Hopefully we can add that soon. We are in the process of adding quite a few new distorts. Anthony can give you more details about what is currently under way and plans for the near term.