Vielen Dank, Malum, für deine Erläuterung! Genau so hat es auch Eric Gerds in einer Mail an mich dargestellt, die ich hier im Wortlaut wiedergebe:
"Harald Bendschneider wrote:
Hallo Herr Gerds,
zum Retuschieren der Nadir- und Zenith-Bereiche von equirectangularen Panoramen verwende ich gern Ihre hilfreichen DOS-Tools.
Nun ist mir mehrfach aufgefallen, dass die rechten und linken Randbereiche der von Pano2Faces erzeugten Cubefaces unscharf verzerrt werden.
Ich ermittele die Größe der zu erzeugenden faces, indem ich die Pixelbreite des von PTGUi gerenderten equrectangularen Panoramas durch 3,14 teile.
Ist das normal und werden diese Randbereiche beim späteren Umformen durch Cube2Pano wieder entzerrt oder ist es , wie ein Forumsmitglied der Panorama-Community meinte, ein Bug?
Zitat: "Da ist ein Bug in der optimalen "Tilesize" drin." Zitat Ende
Vielen Dank für Ihre tollen Tools!
Herzliche Grüße
Harald Bendschneider
You are referring to blurriness at the outer edges of the cube faces. This is NOT a bug with PanoTools. It has to do with mathematics, nothing else.
To fully understand this issue, you need to know what "angular pixel density" means, or pixels per degree.
If I define PPD (pixels per degree) you can see that PPD for an equi pano is PPD = Pano Width / 360. The PPD for an equi pano is a constant throughout the entire image.
However, PPD for a cube face (ie. rectilinear image) is different. The PPD for a rectilinear image changes with distance from the cube face center.
The PPD is larger at the edge of the cube face, and smaller at the center. This is the reason why the edges appear blurry...there are more pixels per degree at the edges than at the center. This is also the reason why VR panoramas on a website have blurry edges, but the center can be sharp.
To preserve all the information (ie. detail) from an equi panorama, we find that:
Cube Face >= 1 + (Pano Width / 3.14)
or approximately:
Cube Face >= (PanoWidth / 3.14)
The previous equation gives us:
PPD at center of cube face >= PPD of equi pano image
[If you want to calculate PPD for yourself, I attach a PPD calclator to this email].
Since all the information is (theoretically) preserved in the cube faces when you use this equation, then converting from cube faces back to equi "should" give you the same equi panorama again.
Of course in reality this may NOT be 100% perfect because of interpolation errors (when you convert from equi to cube to equi, the result is not perfect).
If you are very worried about this issue because you are converting from equi to cube and back to equi again, then I would recommend creating LARGER cube faces.
For example, if you want to patch up the nadir/zenith, then simply use:
Cube Face = PanoWidth/2 (which is >> PanoWidth/3.14)
Then you can convert back to equi pano, and the image should be ok.
Regards,
Eric "