|
As implied from the title, the pano is a second version of the motif shown in pano #9755. Since it was shot in low gegenlicht, I wanted to do a remake, which you can see happened a year later, but only first materialized now.
The view stretches from Nyborg (central eastern Fyn) over Romsø island and Hindsholm peninsula (north eastern Fyn) to northern Fyn. In between some large industrial features and a long hilly area west of Odense. The farthest away features are a transmission tower 63 km away, and the highest hilltop on Fyn 66 km away. The mast is a steel framework construction, so it's thin, and the lens is at it's limit here.
Pano made from 37 LF (RAW), Canon 550D, 300 mm, iso-200, f/6,3, 1/800 sec, developed in DPP (daylight, neutral, moderate sharpness, ALO off, periphal illumination), stitched in PTGui pro, downscaled and sharpened in Irfanview.
Jörg Braukmann, Klaus Brückner, Hans-Jörg Bäuerle, Friedemann Dittrich, Heinz Höra, Martin Kraus, Dieter Leimkötter, Steffen Minack, Jörg Nitz, Danko Rihter, Arne Rönsch, Björn Sothmann, Jens Vischer
|
 |
Comments
LG Jörg
It has been a long process, with the results slowly improving over the years. I now work with PTGui pro, but I started out with Hugin which is a lot like PTGui. I am however not sure if all the features of PTGui is present in Hugin.
First, like you guess, I have quite long overlaps, usually between 33% and 50%. That also gives an advantage if the series jump a little; then only smaller parts needs filling out, unless a crop is made.
Since I made panos of sea horizons early in my "career", the first thing I bitterly learned to do was a fairly straight horizon. Especially Heinz' and HJB1's comments on #7482 helped me along. I don't remember how, but I also early picked up, that apart from the pictures in each end, there only needed to be 2 controlpoints between 2 pictures.
So what I do is to simply throw it in the stitcher, and see what happens. With this type of pano, it often responds by saying that "no" or "only a few" controlpoints are found.
If cp's are found, I always start by checking if there are any with a picture in between, like if picture 2 has cp's on picture 4. If that is the case, I delete them without exception. I then start from an end, making sure that there is 3 or 4 cp's on the first and the last pictures. I magnify to 100%, and find trees, antennas, buildings etc to have controlpoints. No surprise, that absolutely sharpness is needed to put on cp's precisely. A certain amount of contrast in the picture is also very helpful, since it draws the horizon sharper against the sky.
Usually, something presentable comes out from that, but not always. If the pictures - due to slight haze, inability of the photographer, etc - is not 100% sharp, the controlpoints may be a little misplaced. Only quite lately have I learned to check the values of the controlpoints - earlier I simply checked the pano itself, to see if it looked right. Now I check the values, and if the value is more than 1, I manipulate (read "move") the cp with the highest value a little - a pixel left/right/up/down usually tells what direction to take. That way even slightly blurred pictures can be made to fit.
I then choose an anchor image, from the "optimizer" feature, usually in the middle. Should nearer landparts obscure the horizon, I choose the picture with a visible horizon most near the middle.
The last thing is to then set horisontal control points. They are found in the "control points" feature. I have the anchor image to the left, and the first, respectively last, picture to the right. Again I work in 100%, and I put 1 horizontal cp on each pair. I then work my way through. First from the anchor image to each end picture. Then I do the pano and check. If more are needed, I then make a pair of the first and the last picture. Check again. If still more are needed, I make from the anchor image to each picture half way to the end. And check again...
I always shoot with manuel focus, and never with autofocus. I do severel rows in quick succesion, to eliminate the risk of one hopeless frame ruining everything. Luckily I very rarely need that.
Very last thing I do is an optimizing in the "HDR/exposure" feature. I never really needed that when I shot with a crop sensor (like here 1,6), but after switching to full frame I always do that, since transsitions can be very obvious.
Hope that helps!
@Jan, ich bezweifle sehr, daß man mit horizontalen Kontrollpunkten eine Verbesserung erreichen kann. Das Stitchen ist ja eine Optimierungsaufgabe, bei der die Gierwinkel, Nickwinkel und Rollwinkel der Bilder sowie die Brennweite und die Bildverkrümmunsparameter a, b, c so lange gezielt verändert werden, bis eine verfahrensabhängige Summe der Kontrollpunktabweichungen ein Minimum, eben das Optimum erreicht. Es ist mir nicht bekannt, was für ein Optimierungsverfahren PTGui resp. das von Prof Dersch stammende Panorama Tool verwendet. Es sieht aber so aus, das zwischen normalen und horizontalen CP keine Wertung vorgenommen wird. Für einen geraden Horizontverlauf müßten die Abweichungen horizontalen CP aber so gut wie Null werden, während bei normalen CP durchaus noch größere Abweichungen bei einem gefundenen Minimum vorhanden sein können.
U. a. aus diesem Grund habe ich manchmal, wenn ich wenn ich einen besonders geraden Meereshorizont erreichen möchte, es so umständlich wie bei #23846 gemacht.
Es gibt aber auch noch ein rabiates Mittel derart, daß man in einem Bildbearbeitungsprogramm an den Horizont mit Kopier- und Selektierwerkzeugen herangeht.
Dann fällt mir noch der für mich etwas zweifelhafte Hinweis von Jörg Braukmann ein, nach dem PanoramaStudio, das ja keine Kontrollpunkte verwendet, in der Lage ist, supergerade Meereshorizonte zu erzeugen.
First, thank you for still highlightening the old pano. Interesting that you like that better.
Regarding horizontal controlpoints, I tend to agree. I have earlier learned, that well stitched telepanos right from the beginning can be straight. I have however also experienced, that the stitcher can place it too high or too low from the correct horizon, giving a slight curvature. If stitched correct, it can be drawn to a straight position with the mouse, but on long panos that can be difficult to assess when it is truly straight. Then it is easier and better to just place 1 or 2 sets of horizontal controlpoints. LG Jan.
Leave a comment