It's of no importance since all PTGui cares about is the horizontal field of view of the lens (hfov), which is determined by the optimizer. It makes no difference whether you specify the lens as focal length 8mm at crop factor 1.7, 8mm at 1.6, or 9mm at 1.5. PTGui computes an initial hfov from the figures you supply (or from EXIF data in the images), and the optimizer then fine tunes this value.
I used fullframe because the lens is generally regarded as a fullframe lens - i.e. no part of the edge of the image circle of the lens is visible in the frame. However, you can specify the lens as acircular lens and set a circular crop to pass through the corners of the image to use all of the image frame. If the image quality falls off in the corners of the lens, you can set the crop circle smaller so that the corners are excluded.Quote:
Also, you used fullframe for lens type. I used circular.
For the best possible results, it's important to have control points accurately positioned on matching features and nicely spread along the full length of the overlap area. A large number of points is not necessary. The automatic generator will generate as many points as you like (set via Tools->Options), but not all the points may be positioned very accurately and freqently they are not spread evenly either.Quote:
However, after optimizer I got max distance 15.xxx
Poorly positioned points will usually be characterised by large distance values in the control points table after optimization. These points should be either deleted or manually repositioned accurately. You can use the Delete Worst Control Points option on the Control Points menu to delete all the really bad points in one fell swoop, but PTGui will not remove a point if by doing so less than 7 points would remain for pair of images.
Removing bad points automatically may remove points that have every appearance of being accurately positioned. This can happen when features are affected by parallax or when the lens parameters are not yet right. To get the lens parameters right, you need a nice spread of control points. In simple terms, if you want the images to align well all along the length of the seams, you need control points placed all along the seams. So it pays to check the spread of points between pairs of images visually and add a few points manually where they are lacking. It is all too easy to get a "very good" optimizer report when points are clustered in the middle of the image overlaps, but the images may still not align well in the corners where there are no control points owing to the lens distortion not being adequately corrected. Adding points towards the corners will help the optimizer to correct the distortion there and get good alignment.