* A suitable beefy browser equipped machine for display, depending on image sizes and ALTO complexity
*[deepzoom](http://search.cpan.org/~drrho/Graphics-DZI-0.05/script/deepzoom)(only if tile based display is enabled)
* Perl (required by deepzoom)
## Potential improvements
...
...
@@ -30,17 +32,17 @@ An enhanced [ALTO](http://www.loc.gov/standards/alto/)-viewer for Quality Assura
* Optional tile mode instead of a single image for display
* Optional removal of destination files when source files are removed
* More flexibility and customization (quack was developed for internal use at the State and University Library, Denmark)
* ALTO files at another location than image files
* Add image & ALTO sample files
* Integrate greyscale statistics
* Show blown high- and low-lights on the thumbnail in folder view
## Verbiage
This is basically a simple bash script that grew to 600+ lines. It works on a collection of images with corresponding ALTO-files with OCR and segmentation markup. It creates a HTML page for each image and uses the relevant parts of the ALTO files for creatint TextBlock overlays with OCR inspection. No webserver is required as the pages can be used directly from the file system.
This is basically a simple bash script that grew to 600+ lines. It works on a collection of images with corresponding ALTO-files with OCR and segmentation markup. It creates a HTML page for each image and uses the relevant parts of the ALTO files for creating TextBlock overlays with OCR inspection. No webserver is required as the pages can be used directly from the file system.
Currently the display of the scans is not tile based and thus requires the browser to handle the full images directly. This can be quite taxing. It has been tested with Firefox & Chrome. Chrome was markedly faster as of 2013-09-17.
It can be quite heavy for the browser to display large images with OpenSeadragon, when there is a lot of ALTO-derived overlays. This is especially true if tiling is not used. Display has been tested with Firefox & Chrome. Chrome was markedly faster as of 2013-09-17.
The choice of PNG was due to the QA-focus - no JPEG artifacts, thanks! This should probably be more flexible as the input can be JPEG (which makes the conversion to PNG plain silly) and because some people do not require pixel perfectness for their QA.