I think it is sensible to interpolate lengths in the reciprocal space. This means that LensFun should not use focal length f and distance d, but the reciprocal values 1/f and 1/d. The algorithms would not change much: You just need to substitude f and d by 1/f and 1/d in both the function parameters and the XML file parameters before the actual interpolation.
I did some measurements recently, and in reciprocal space, the curves are *much* more linear than in normal space. This means the interpolation would work much better. And then: How are the users supposed to code a focal distance of infinity? In reciprocal space, it doesn't matter really whether the users codes infinity as d=100, d=1000, or d=1000000. However in the current implementation, LensFun will sonn fail to apply Hermite splines through the points. ------------------------------------------------------------------------------ Put Bad Developers to Shame Dominate Development with Jenkins Continuous Integration Continuously Automate Build, Test & Deployment Start a new project now. Try Jenkins in the cloud. http://p.sf.net/sfu/13600_Cloudbees _______________________________________________ Lensfun-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/lensfun-users
