 |
|
|
|
|
|
 |
| ||||||||
 |
 
|
|
|
|
|
|
|
|
|
| ||||||||
|
|
| 
| User list | Rules | You are not logged in.
Pages: 1
Don't understand Output Size at 100%, because you sad that it's calculated by the highest Focal length in the panorama.
So in this Panorama the highest focal lenth is 140,67 mm at 35mm(36x24mm). See picture.
And the Panorama is 360°.
And the output resolution from the Camera is 4416 Pixel width.
So I calculated:
2 * arctan(36 / (2 * 140,67)) = 14,58377606°
360° / 14,58377606° = 24,68496489°
24,68496489° * 4416 Pixel = 109.008,8049 Pixel
And like you see on the screenshot, when I go to 100% I will get 113.277 Pixel rendersize and not about 109.009 Pixel.
Why is that?
Maybe the problem is that the focal lenth is different from Fov.
140,67 mm focal lenth should be 14,58377606° and not 14,028° Fov. See last picture.
Last edited by Sam (2010-11-24 18:33:01)
Offline
Sam wrote:
140,67 mm
The precision (1/100 mm!) implied by this value is completely out of reach, the actual focal length was probably larger than 141 mm or smaller than 140 mm... (Focal length values referred by manufacturers are "commercial values", not actual values. As a rule, the recorded EXIF values are inaccurate.)
Many zoom lenses are to be handled with great care to avoid small focal length changes between shots.
Sam wrote:
360° / 14,58377606° = 24,68496489°
In this computation source images overlap should be taken into account.
Last edited by GURL (2010-11-23 19:32:08)
Offline
I'm really troubled by Sam's obsession with mathematics. What does it matter if the pano is 113K or 109K pixel wide?! If you're worried the image is "too big to be true" - scale it down to whatever you think is appropriate. Why do you disregard the fact that the images get wrapped (transformed) and in that process their sizes change?
Also, as Georges said, there's no such precision in lens and EXIF files.
Offline
[bo] wrote:
I'm really troubled by Sam's obsession with mathematics.
To understand how things works mathematics are often a powerful tool!
In this case Sam is troubled because results don't correspond to his theory (theory which has to be refined: precision and overlap must be taken into account.)
Offline
To understand how things works mathematics are often a powerful tool!
Thank you.
I think everybody knows that behind this programm there is a lot of mathematic.
And everybody can try for yourself that with a calculatur
140,67 mm focal lenth should be 14,58377606° and not 14,028° Fov.
There is somethink wrong. An it is only about 3%, but maybe it has some bad effect, we don't know.
I hope they perceived this post.
Offline
For me it would be convenient, if the render output size, in preferences, could not be specified in percentage, but in pixels (wide).
Because I always want to have the same output size. For me, 108 544 pixels wide. Because I take the photos almost with the same setting. The 108 544 pixels, I calculated by me self, because I think that your calculation is not accurate. Perhaps you have a different formula to calculate the Fov. And therefore the differences.
It would be very interesting how you calculated the Fov.
Thanks
Offline
Sam wrote:
For me it would be convenient, if the render output size, in preferences, could not be specified in percentage, but in pixels (wide).
You can enter the size (width) in pixels in the topmost leftmost field of the Render window ...provided it's not larger than the computed "100%" width (see screen capture) but you can't specify that in global preferences.
Sam wrote:
Perhaps you have a different formula to calculate the Fov. And therefore the differences.
It would be very interesting how you calculated the Fov.
Computed FOV depends on:
- FOV values as computed by the optimizer for each source images
- yaw values computed by the optimizer for the leftmost and for the rightmost source images
- number of pixels per angular unit at the image center (wich depends on actual focal length, distance settings, pixel width of the sensor and actual sensor size)
- selected projection mode
- image crop (automatic or manual)
A difference between the FOV you compute and the FOV autopano computes is unavoidable but it should not be very large...
Last edited by GURL (2010-12-30 13:41:45)
Offline
GURL wrote:
You can enter the size (width) in pixels in the topmost leftmost field of the Render window ...provided it's not larger than the computed "100%" width (see screen capture) but you can't specify that in global preferences.
I meant global preferences, so that I can use Autorender.
A difference between the FOV you compute and the FOV autopano computes is unavoidable but it should not be very large...
Why is it unavoidable?
In my formula 140.19 mm focal lenth should be Fov (horizontal) 14.63316964 ° and not 14.076 °. See screenshot of APP.
To calculate the Field of view, without a calculater you can use. It looks like, they are calculated like my results.
http://www.howardedin.com/articles/fov.html
http://www.tawbaware.com/maxlyons/calc.htm
http://www.bobatkins.com/photography/te … _view.html
I don't think that the roll value of 0.373 chance the FOV so match.
The difference is 0.557169644 degrees between me and the APP calculation.
That makes at 39 columns like I use, at 140mm focal length, 21.72961612 degrees difference and that's a lot.
Last edited by Sam (2010-12-30 17:16:57)
Offline
I agree that output size specified in pixels in the prefs. would be good, but it leads to a problem:
- specification in percents works for any pano
- specification in number of pixels (width ? height ? both ?) will work only for dedicated panoramas, shot with a specific focal lens, number of shots, etc...
Regarding the difference between your calculations and the results given by APP, I suspect that the difference comes from inner adjustments made by the software when it calculates the stitching between the images. What do Kolor say about that ?
Offline
The final panorama size calculation is done with 2 aspects in mind :
- the longest focal length is used as reference for the 1:1 ratio, one pixel in input, one in output
- when using projection mode other than spherical, this rule is applied on the (0,0) coordinate : viewpoint center.
Note : the focal length used is not the one provided with exif, it's the once calculated by the software ( it can vary a little from the one in exif ). See the layer editor learn about the calculated values.
Offline
Pages: 1
