OCR Shop XTR 5.5.0 Release Notes July 7, 2004 For additional information, please refer to ocrxtr_README.txt and the OCR Shop XTR manual. Both are installed along with OCR Shop XTR in /opt/Vividata/docs and the manual is available at www.vividata.com under Customer Service. For technical questions or purchase information, contact Vividata: * Click Contact Support at www.vividata.com * Call 1-510-658-6587 =============================================================================== Contents: * New Documentation * Major Updates and Bug Fixes * Hidden Options * Known Issues and Usage Notes * Major Updates from Previous Releases =============================================================================== =============================================================================== New Documentation OCR Shop XTR now comes with a Quickstart guide to help you install and start using OCR Shop XTR. Please find the Quickstart guide in /opt/Vividata/docs/ocrxtr_README.txt. New documentation files included /opt/Vividata/docs: ocrxtr_README.txt Installation and usage information ocrxtr_release_notes.txt Release notes and known issues ocrxtr_tips.txt Tips for improving OCR results and performance To create output containing detailed content and structural information, along with confidence values, font metrics, and formatting: ocrxtr_xdoc.txt Instructions for using the XDOC format core12xdc.pdf Scansoft XDOC documentation kdoctext.h Supplemental XDOC information from Scansoft For forms processing and custom document segmentation: ocrxtr_rdiff.txt Introduction to forms processing with OCR Shop XTR rdiff_sample/ Files to accompany the rdiff example New sample input images included in /opt/Vividata/bin: cyrillic_with_english.tif letter_24bit.pdf letter_24bit.tif letter.tif low_resolution.tif french_german.tif letter_24bit.psd letter.pdf low_contrast.tif =============================================================================== Major Updates and Bug Fixes New features and behavior changes --------------------------------- * On installation, the user may now set VV_HOME to any directory in order to install OCR Shop XTR there. The default installation directory is /opt/Vividata. In addition, a non-root user may install OCR Shop XTR, if VV_HOME is set to a directory for which the user has read/write permission. * OCR Shop XTR no longer requires the user to set VV_HOME when running the software. A user is only required to set VV_HOME if OCR Shop XTR is installed in a directory other than the default /opt/Vividata. * The installer, license manager, and OCR Shop XTR application log and error messages have been expanded and improved. * OCR Shop XTR now ignores the TIFF fillorder bit by default when a TIFF image is used as input. If the user needs OCR Shop XTR to obey the TIFF fillorder bit, the environment variable "VV_IGNORE_FILLORDER" should be set to "n", "no", or "0", or the command-line parameter "ignore_tiff_fillorder" should be set to "n", "no", or "0". Please see the OCR Shop XTR manual for more information. In previous versions of OCR Shop XTR, the TIFF fillorder bit was obeyed by default. OCR Shop XTR now ignores the TIFF fillorder bit by default, because many TIFF writers set the fillorder bit incorrectly. If an incorrectly set TIFF fillorder bit is obeyed when OCR Shop XTR reads a TIFF input file, the image data will not be recognizable as text. * A signal handler now catches unexpected crashes in OCR Shop XTR, which enables OCR Shop XTR to return the license token and delete temporary files after a crash. Previously, if OCR Shop XTR crashed, the user would need to reset the license manager by hand; this is no longer neccessary. * Added a new command-line option "out_depth" to specify the bit depth of any output image data. This affects PDF, HTML, and graphics output. This option provides more control to the user over the filesize of PDF, HTML, and graphics output. For PDF and PS input, the default "out_depth" is 1 bit-per-pixel. For all other input image formats, the default "out_depth" is the bit depth of the input image. * Improved the default processing time and memory usage for input PDF and PS files, and provided the user with control over the processing time and memory usage for PDF and PS input. The memory usage and processing times associated with a PDF or PS input file are most affected by the bit depth at which the input file is rendered. By default, PDF and PS input files are now rendered at 1-bit per pixel. Previously, they were rendered at 24-bits per pixel, resulting in a much larger amount of image data to store and process, and potentially causing memory or swap file problems for large, multipage input PDF and PS files. Now, the bit depth at which PDF and PS input files are rendered is based on the setting of the "out_depth" option, so that OCR Shop XTR retains only as much image depth information as need in output PDF, HTML, or graphics. When creating PDF, HTML, or graphics output, the user should set "out_depth" to 8 or 24 if they wish to preserve 8 or 24-bit image data from an input PDF or PS file, keeping in mind how it might affect memory usage and processing times. If "out_depth" is not set, the default output bit depth will be 1, even if the input PDF or PS file contained 8 or 24-bit image data. The "out_depth" setting affects only how PDF and PS input files are rendered and the appearance of output image data. It does not affect the recognition quality, because the OCR engine only processes 1-bit image data. Note: The default "out_depth" when using other types of input images (JPEG, TIFF, etc) is the bit depth of the input image. * The default for the output PDF format has been changed to "img_text" from "normal", because "img_text," where the full page original image appears with the invisible recognized text underneath, is most commonly used. * Improved error messages if the license log file can not be opened. * The resolution at which the input image is interpreted (controlled manually by the "in_res" option) now affects the appearance of PDF output. * A new command-line option, "timeout", was added to provide the user with the ability to set a timer in seconds for recognition. Bug fixes --------- * The image data of input images with non-square resolutions, such as faxes, was rotated too much when deskewed for PDF, HTML, or graphics output, affecting the appearance of these output formats, but not the OCR results. Now the non-square input images are deskewed the correct amount in graphical output. * A bug fix eliminates a periodic crash bug that occurred on creation of full page graphics output using the "output_whole_image" flag, if the image was deskewed. * When the input image was deskewed, any PDF, HTML, or graphics output was converted to a bit-depth of 24 bits per pixel. Now, the output bit-depth will match the input image's bit-depth. The format of the output image data will also affect the output bit-depth used, because, for instance, the JPEG format does not support a bit-depth of 1. In addition, the output bit-depth is also affected by the "out_depth" command-line option. The user should be aware that PDF and PS input files are considered to have a bit-depth of 1. * For input images with a bit-depth of 8 or 24 bits per pixel, OCR Shop XTR could not create valid output graphics files in the tiff-g31d, tiff-g32d, or tiff-g42d formats. OCR Shop XTR now creates valid output graphics files in these formats in all cases. * Fixed a "hidden" option, whole_page_image, to allow user to generate a full page image graphic. * A segmentation fault occured when a text file was passed to OCR Shop XTR instead of an image file. * Input PDF files with filenames that contain spaces are now processed successfully. * OCR Shop XTR would finish without creating an output file or would crash when the user did not have permission to write the output file, temp files, or log files. OCR Shop XTR now quits gracefully with a coherent error message when the user does not have write permission. * OCR Shop XTR did not always check whether the user's license permitted PDF and PS input. Now when the the user sends a PDF or PS file as input, OCR Shop XTR always checks to ensure that PDF/PS input is licensed. If PDF/PS input is not licensed, it will print an error message. * Automatic orientation correction worked only on the first page of a multipage output PDF document. Now, OCR Shop XTR correctly rotates every page of an output PDF document. * In some cases, images were misplaced in PDF output or OCR Shop XTR crashed while creating the PDF output. Images are now placed correctly in PDF output and OCR Shop XTR should not crash. * A bus error (signal 10) occurred on some systems on exit when VV_HOME was set. This bus error no longer occurs. * Processing of certain image files in particular circumstances caused OCR Shop XTR to crash. Now OCR Shop XTR processes these files successfully. * Images were sometimes not placed correctly in PDF "normal" output. =============================================================================== Hidden Options These options are available for debugging purposes, but are not officially supported and do not appear in the OCR Shop XTR documentation. * "-debug=1" instructs ocrxtr to write verbose debug output to the console. Turn on this option if you encounter a problem, and wish to report it to Vividata. * "-output_whole_image=y" allows the user to create graphics output of the entire input image. * "-out_debug_files=y" will have ocrxtr create two output files for debug purposes only: unconv_input_file and converted_input_file. "unconv_input_file" is created after the input file is read, placed in an internal data structure, and written back out as a TIFF image file. "converted_input_file" is created after ocrxtr converts the input image data to 1-bit image data prior to recognition; this output TIFF show what data is sent to the OCR engine for processing. Both of these files are multipage TIFF files; ocrxtr appends to them on subsequent runs, rather than overwriting them. =============================================================================== Known Issues and Usage Notes * The "remove_halftone" option is not functional at this time. * For GIF input files, the image dpi stored in the header is ignored in favor of the default resolution of 300x300. If the in_res option is set, it will take precedence, as expected. * With XDocPlus output, a temp file called "facp.@fa" may be created in the working directory and not cleaned up when OCR Shop XTR finishes. * Using a binary file in place of a text file will probably cause a bus error and OCR Shop XTR will crash. Setting any of these command-line options to the name of a binary file will trigger this bug: image_list, image_rdiff_list, user_lexicon, read_params. * Deskewing is very slow for large image output, either through embedded document output or direct image output. Deskewing occurs when the input image is crooked and must be rotated slightly for the engine to properly recognize the text; it may be turned off with the parameter "-deskew=N". Image output with deskewing is only really slow when the image being output is about the same size as the original input document. Outputting small image regions is reasonably fast. * The user lexicon, defined by the file named in the "user_lexicon" option, and the character set, defined by the "char_set" option, may contain ASCII characters, interpreted with respect to the loaded code page. The user lexicon and character do not support non-ASCII characters. =============================================================================== Major Updates from Previous Releases Release 5.1 * OCR Shop XTR now uses a new command-line based installer. * OCR Shop XTR also now uses a new proprietary license manager. Old FlexLM license keys will not work with the new license manager; please contact Vividata if you have questions about upgrading. http://www.vividata.com * PDF output has been upgraded, with improved compression for smaller output filesizes and improved operation. * The conversion of 8 and 24-bit images to 1-bit has been updated to provide better OCR results. =============================================================================== ===============================================================================