ArtScan

[qwer]

About 3 month ago started project of scriptable scans processor http://code.google.com/p/artscan/ , written in Python/PyQt4.

Filters/transformations easily added through plugins (example - rotation plugin http://code.google.com/p/artscan/source ... tion_qt.py ).

To download sources you need Mercurial: http://mercurial.selenic.com/wiki/Downl ... ryPackages , download command : hg clone https://artscan.googlecode.com/hg/ artscan

To run you need Python (tested with version 2.5): http://www.python.org/download/releases/2.5.4/
PyQt4 for Python 2.5 : http://www.riverbankcomputing.co.uk/sof ... t/download
additional packages : Babel : http://babel.edgewall.org/wiki/Download

Every plugin is descendant class of Plugin.
In constructor __init__ first filled plugin description fields (shown to the user in plugins tree and parameters editor), and added plugin parameters descriptors (first added 3 standard parameters for document/page/layer selection, and additional - rotational angle, rotational quality. Each parameter has short tag (by which you get access to parameter value), full localized name and description name,descr (displayed to user), default value (val_def), by it's type determined parameter type (int/float/string/bool), additinal parameter options (minimum, maximum, step etc), vals - list of possible values.
Actual plugin processing in done in method 'execute'. It receives values of entered parameters when user click on 'Apply' button in parameters editor.
Method self.getLayersSelection of Plugin object returns list of selected documents/pages/layers objects based on values of parameters 'documents', 'pages', 'layers'.
All selected layers objects processed in loop: for layer in self.getLayersSelection(params):.
First, image of layer readed : image = layer.getImage(), then done transformation of image : image.transformed (look in Qt/PyQt QImage documentation on possible tranformations). Then new image saved for current processed layer instead of old, and page view refreshed (layer.setImage(image, rebuildView=True)).

Currently developing/testing/planned : backround plugins execution, undo for layers/zones operations, rectangular/polygonal zones creation/editing, plugins for effects/transformations (using Qt, PIL, or PythonMagick - ImageMagick binding) / djvu,pdf import/export (split Djvu layers editing possible) / scanning (through sane/twain, with control of automatic scanners) / OCR integration (ocropus,cuneiform or other engines) and text layer editor.
Selective pages/layers/zones processing is possible by adding attributes for that objects ('Attributes' parameter in document/page/layer/zone properties editors tabs above pages list).
Also possible immediate scripts execution (in script editor on plugins panel). Executed script commands shown in log text area (bottom of plugins panel), user can copy them to script editor, and make own script by combining many processing commands.

Class hierarchy:

Code: Select all

ArtScanApplication - main application
  ArtScanMainWindow - main window
    Documents - container of opened documents and tab widget
      Document - document, container of document pages and page view area
        Page - represents document page with multiple image layers
          Layer - represents page layer
            Zone - selected zone on layer
    Plugins - container of loaded plugins
      PluginsGroup - group of plugins in plugins tree
      Plugin - plugin 

[daniel_reetz]

Hey Qwer, nice work! And thanks for the technical introduction.

Is ArtScan primarily set up for the output of flatbed scanners, or does it accommodate camera scans?

[daniel_reetz]

I see in the right hand panel there is an "OCR" selection. Is that implemented, or just a placeholder?

[qwer]

Is ArtScan primarily set up for the output of flatbed scanners, or does it accommodate camera scans?

no difference, it depends only on developed transformation/filtration plugins and their combination.

I see in the right hand panel there is an "OCR" selection. Is that implemented, or just a placeholder?

Placeholder, I plan to use ocropus/cuneiform.

[qwer]

Update:
- added parameters validation (marks wrong parameters by red color),
- added couple of plugins using Imagemagick (rotate, despeckle, sharpen), works good but slow, PNG format much slower, use TIFF instead - its faster and preserves original compression. Strange but despeckle gives very weak results, and has no execution parameters to tune it up.

[qwer]

For testing purposes added GraphicsMagick command-line contrast enhancement plugin. Seems to process images much faster than ImageMagick (IM loads image up to 10 seconds).
http://code.google.com/p/artscan/source ... rast_gm.py

[qwer]

Completed basic zones operations:

[rob]

Wow, it's looking very nice!

I hope you can find some kind of page uncurling library :/

[qwer]

I hope you can find some kind of page uncurling library

I doubt, perhaps I'll have to write it myself.
BTW have you received satisfactory results with your algorithm ?

[rob]

Only for mildly curled pages. For severely curled pages, the algorithm results in a page full of fail