Dear all, I have some remarks and questions about resampling using OTB, in particular using RigidTransformResample application scheme (which is based on a StreamingResampleImage filter).
let's show you my problem : apply a resampling of an input with a factor 10 : otbcli_RigidTransformResample -in ~/workspace/ORFEO-TOOLBOX/src/OTB-Data/Input/QB_Toulouse_Ortho_PAN_WGS84.tif -out /media/otbnas/jguinet/REECH_OTB/QB_Toulouse_Ortho_PAN_WGS84_scale_0.1.tif -transform.type id -transform.type.id.scalex 0.1 -transform.type.id.scaley 0.1 -interpolator bco then open input and output with gdal : we observe a shift between resampled data and input one. Resampling is done as follow : project each pixel (integer) of output c.s to pixel position (fractional) in input c.s. interpolation is then used to approximate data. - use output information (origin,spacing,size) to compute output information (the output grid) - create the deformation field (the distance between physical position of each point input grid and point in output space) using a trasnform set to identity. - interpolate value of input position from output one using deformation field. using this scheme and a WGS84 input for example, each geodetic position of an output pixel is mapped with the same postion in input for example origin of output is interpolated at origin of input. Using a 0.5 scaling and linear interpolator it gives the result depicted in the illustration "Resampling.jpg" : (resampled image is depicted with dots line, resampled data is interpolated using dashed input pixel (linear interpolator use only dashed pixel, and is nearly the same as a NN interpolator) ). Thus shift is observed. Considering the (0,0) pixel the resampling strategy can be illustrated as "Origin_Resampling.jpeg" if we consider that a pixel is a point convolued by a PSF, i think that an upsampling should be as illustrated in "New_Resampling.jpeg" I wonder if the transform should be a mid spacing shift (offset=(outputSpacing-inputSpacing)/2). using this transform offset results in resampled data exactly superimposed using gdal. moreover using this scheme the new origin might be shifted with outputSpacing/2 (with the drawback of footprint change between output and input at origin). What do you think about that ? Thanks Jonathan -- -- Check the OTB FAQ at http://www.orfeo-toolbox.org/FAQ.html You received this message because you are subscribed to the Google Groups "otb-users" group. 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/otb-users?hl=en --- You received this message because you are subscribed to the Google Groups "otb-users" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
<<attachment: Origin_Resampling.jpeg>>
<<attachment: Resampling.jpeg>>
<<attachment: New_Resampling.jpeg>>
