[gscan2pdf-announce] scantpaper v3.0.24 released

[Jeff <jf...@po...>]

Changes compared to v3.0.23:

* OCR text is now aligned to its baseline instead of being vertically
   centred in each bounding box, so words with and without descenders
   line up correctly with the scanned image.
* hOCR font metrics (x_size, x_height, x_ascenders, x_descenders,
   x_fsize) are now preserved when loading and saving documents.
* A horizontal or vertical split view is now shown on startup and the
   imported image appears immediately, instead of the split panes never
   being attached (the tabbed view stayed in place) so the image only
   appeared after cycling through the view modes. The split divider's
   position is remembered across sessions.
* "Clear OCR" no longer fails with an internal error. It now removes the
   OCR text layer from every selected page, saves the change so it
   survives a reload, and makes clearing several pages at once a single
   undoable step (one Undo restores all of them).
* Saving a PDF with "Automatic" compression now produces much smaller
   files for the same scan: bilevel pages are encoded as CCITT Group 4,
   pages with transparency use lossless Flate, and colour/grayscale pages
   use JPEG.
* The JPEG quality setting in the Save dialog now affects PDF output, so
   lowering it reduces the file size. An imported JPEG saved with
   "Automatic" compression is still embedded unchanged.
* The Save dialog gained two new PDF compression choices: "JPEG 2000"
   for smaller continuous-tone images and, when the `jbig2` program is
   installed, "JBIG2" for roughly half-size bilevel (black-and-white)
   pages. Both are opt-in; "Automatic" behaves as before.
* A new "Bilevel compression" preference (Edit → Preferences) chooses
   the codec "Automatic" uses for black-and-white pages: CCITT Group 4
   (the default) or JBIG2 when the encoder is available. It leaves colour
   and transparent pages unaffected.
* The PDF "Compression" menu no longer offers "LZW" or "G3". Both were
   misleading: "LZW" always produced the same Flate output as the "Flate"
   option, and "G3" produced CCITT Group 4 like the "G4" option. A saved
   preference for one of the removed options keeps its output (Flate or
   Group 4) automatically.
* Text you add manually to the OCR text layer (drawing a rectangle and
   typing a word) is now included in the text layer of the saved PDF.
   Previously such manually added words were silently dropped, so the
   saved PDF's searchable text contained only the automatic OCR output.
* You can now draw the first selection rectangle in a layer pane (text
   layer or annotations) on a page that has not been OCRed, or whose
   layer is empty, so a text layer can be started by hand before OCR is
   ever run. Previously drawing was disabled until the pane already had a
   layer, which only OCR (or a prior slice) could provide.
* A focused OCR text slice can now be rotated 90 degrees clockwise or
   anticlockwise from the text layer controls, so vertical text (e.g.
   rotated index columns) that tesseract read wrongly can be authored by
   hand: the corrected glyphs are drawn and saved at the rotated angle,
   and the box itself stays axis-aligned. Slices added while a slice is
   rotated inherit that rotation, letting you type a run of vertical text
   in one go.
* Reversed the order of the rotate controls, both for the page rotate
   icons in the toolbar and for the text in the layer editor.
* A word you add to the OCR text layer is now oriented automatically
   from the rectangle you draw: a tall, narrow selection becomes vertical
   text and a wide, short one stays upright, so a box drawn for a
   vertical column renders legibly instead of as shrunken horizontal
   text. This also applies to the first slice that creates a new text
   layer and to duplicated slices, unless a rotation is already
   inherited.
* The selection rectangle now stays aligned between the image and text
   windows in a split view at every zoom level, and when the scan has a
   non-square resolution or the display uses HiDPI scaling. Previously
   the rectangle in the text window drifted out of alignment as you
   zoomed in or out.
* The text layer's sort criteria now takes effect as soon as you enter
   the editor. Previously it was shown as "Sort by confidence" but
   navigation actually went by position until you changed the selection
   and changed it back; now the selected order is applied immediately, so
   the first entry navigates the way the control says.
* Accepting a correction in the text layer editor now moves on to the
   next slice in the current sort order, instead of staying on the one
   you just corrected. In a confidence-sorted pass this walks from the
   least to the most confident text automatically, so you no longer have
   to re-select the next slice after every fix. Correcting the final
   slice keeps it focused.

Available on Github:

https://github.com/carygravel/scantpaper/releases/tag/v3.0.24

Or via the PPA:

https://launchpad.net/~jeffreyratcliffe/+archive/ubuntu/ppa