In my experience, the sigmoid tends to stay in the limits of the first schema in your pdf. A very small deviation has a huge impact so that if you are not looking for strange-looking results, you tend to stay close to the diagonal, especially with jpegs where doing too much correction triggers banding. For example, here are the 3 curves I used on o1 for lowering contrast, increasing contrast, and the third D-shaped curve which gives the results closest to faux-bracketing.
2012/12/21, JohnPW <[email protected]>: > I mean that as the contrast factor increases, the curve trends to be more > like a step function (where all pixels to the left of the midpoint become > black and all to the right become white.) > Attached a horrible pdf drawing (sorry I had no handy vector drawing tool > and had to draw with the trackpad too!) > > I assume the inflection point of the sigmoid curve only shifts along the > diagonal as the contrast factor is changed. Is this correct? If so, at 100% > > the invlection point is at the white point and the slope of the curve > coming out of the white point is is very steep (essentially straight down. > If instead, the inflection point stays at the same level (only moves left > and right,) it would be even steeper. No, I guess this is impossible as the > > curve would emerge from the middle of the right side of the graph (and the > white point would not be anchored at the top right.) Yup, I'm just letting > my thoughts wander here. :-) > > Anyway, with such a steep curve coming out of the white point, surely this > would effectively blow out the highest of the highlights a good way down > the scale, wouldn't it? Which would make details in the extremes disappear. > > John > > > On Thursday, December 20, 2012 3:47:51 PM UTC-6, Frederic Da Vitoria wrote: >> >> Apart from the "and a sigmoidal (s) shape between perfectly straight and >> infinitely steep (essentially a step function)" (my English is not good >> enough to understand your meaning, so I can't say if this is correct or >> not), yes, this is correct. In the Gimp curves filter, the sigmoidal curve >> >> could be inflected as a printed "S", in which case the results are >> accentuated (by increasing the dynamic range in the middle tones and >> losing >> details in the extremes), or like a rounded "Z" and this results >> subjectively in a loss of contrast but should improve the details in the >> extremes. >> >> 2012/12/20 JohnPW <[email protected] <javascript:>> >> >>> I'm glad this is stimulating discussion. This is the kind of thing I >>> enjoy reading here (especially when somebody can explain things they >>> know.) >>> >>> Anyway, I'd like to figure out some things about sigmoidal contrast. Is a >>> >>> "sigmoidal contrast" simply an approach to setting the contrast curve we >>> >>> all know from photography and digital imaging? In other words, is it a >>> subset of possible contrast curves that could be uses which follow a >>> sigmoidal shape as described in the formula? >>> From what I can understand, this means it is a typical contrast curve >>> but: >>> • with the endpoints locked down at blackest black and and whitest white >>> • with a single inflection point somewhere in the middle >>> • and a sigmoidal (s) shape between perfectly straight and infinitely >>> steep (essentially a step function) >>> • presumably this function always has a positive slope and the 's' shape >>> >>> is never inverted into a backward 's' shape. >>> Are my assumptions correct? >>> >>> >>> >>> On Wednesday, December 19, 2012 5:09:02 PM UTC-6, Bruno Postle wrote: >>>> >>>> On Wed 19-Dec-2012 at 12:13 +0100, Frederic Da Vitoria wrote: >>>> > >>>> > But now that I think of it more closely, I understand that >>>> > JohnPW's question is still unanswered and that my answers >>>> > completely missed the point. I expect the faux-bracketing to keep >>>> > the lightest parts of the darkest exposure and the darkest parts >>>> > of the lightest exposure. >>>> >>>> I'm not entirely sure what you are trying to do, but Hugin will >>>> extract 'bracketed' exposures from any photo. By default it uses a >>>> 'sigmoidal' camera response curve to map the data to linear space, >>>> multiply and then map back again - This curve will be quite accurate >>>> if you have calibrated your camera photometric parameters. >>>> >>>> All you have to do is load your photo and increment Eev for the >>>> input or output. >>>> >>>> You can then enfuse these brackets, but this process only really >>>> makes sense if you start with 16bit per channel data created from >>>> RAW. >>>> >>>> -- >>>> Bruno >>>> >>> -- >>> You received this message because you are subscribed to the Google Groups >>> >>> "Hugin and other free panoramic software" group. >>> A list of frequently asked questions is available at: >>> http://wiki.panotools.org/Hugin_FAQ >>> To post to this group, send email to >>> [email protected]<javascript:> >>> To unsubscribe from this group, send email to >>> [email protected] <javascript:> >>> For more options, visit this group at >>> http://groups.google.com/group/hugin-ptx >>> >> >> >> >> -- >> Frederic Da Vitoria >> (davitof) >> >> Membre de l'April - « promouvoir et défendre le logiciel libre » - >> http://www.april.org >> > > -- > You received this message because you are subscribed to the Google Groups > "Hugin and other free panoramic software" group. > A list of frequently asked questions is available at: > http://wiki.panotools.org/Hugin_FAQ > To post to this group, send email to [email protected] > To unsubscribe from this group, send email to > [email protected] > For more options, visit this group at > http://groups.google.com/group/hugin-ptx > -- Frederic Da Vitoria (davitof) Membre de l'April - « promouvoir et défendre le logiciel libre » - http://www.april.org -- You received this message because you are subscribed to the Google Groups "Hugin and other free panoramic software" group. A list of frequently asked questions is available at: http://wiki.panotools.org/Hugin_FAQ To post to this group, send email to [email protected] To unsubscribe from this group, send email to [email protected] For more options, visit this group at http://groups.google.com/group/hugin-ptx
<<attachment: 2012-12-21_131719.png>>
<<attachment: 2012-12-21_131819.png>>
<<attachment: 2012-12-21_131913.png>>
