Currently, any information on image resolution provided in the original image (and made available via OcrdExif in Workspace.image_from_page) is ignored when saving derived images in the workspace (via Workspace.save_image_file). Due to PIL.Image format internals, the PNG then contains a setting of 72 DPI however. This might create problems for processors that look at the derived image files alone.
But this is hard to fix in core: the image passed to save_image_file could come from anywhere (and usually does not have an info['dpi']; even simple PIL.Image operations omit that in the result).
Realistically though, it will have been created some way from the source image file under the same pageId, and since rescaling is currently not permitted in the spec, one could assume the same DPI for all derived images.
Currently, any information on image resolution provided in the original image (and made available via
OcrdExifinWorkspace.image_from_page) is ignored when saving derived images in the workspace (viaWorkspace.save_image_file). Due to PIL.Image format internals, the PNG then contains a setting of 72 DPI however. This might create problems for processors that look at the derived image files alone.But this is hard to fix in core: the image passed to
save_image_filecould come from anywhere (and usually does not have aninfo['dpi']; even simple PIL.Image operations omit that in the result).Realistically though, it will have been created some way from the source image file under the same
pageId, and since rescaling is currently not permitted in the spec, one could assume the same DPI for all derived images.