Piper's Pit BBS/FTP: ibm 0010

home *** CD-ROM | disk | FTP | other *** search

/ Piper's Pit BBS/FTP: ibm 0010 - 0019 / ibm0010-0019 / ibm0010.tar / ibm0010 / PROD_OCR.ZIP / README.DOC < prev next >

Wrap

Text File | 1989-05-26 | 23.1 KB | 555 lines

GENIUS GS2000 - GS4000 PRODIGY OCR V1.69 (c) 1988, 1989 Synergy (UK) Ltd. All rights reserved. 25th May 1989 NEW FEATURES --- MAKE SURE YOU READ THIS SECTION CAREFULLY --- ============ V1.67+ :- 1) Context checking to reduce the chances of 0Oo and 1lI characters occurring out of context. 2) Context checking not enabled in learn mode, so upper/lower case characters can occur unexepectedly in learn mode. THESE ONLY NEED TO BE TAUGHT IF THE REAL CHARACTER IS NOT SHOWN IN THE POSSIBLE CHARACTER STRING. --- 3) The Font dictionary is saved to disk when overflow space has been used up in learn mode. It is then rebuilt into memory. This shows up as a pause when you are teaching. V1.69+ :- 4) Pixel display in learn mode. 5) Font status on display in learn mode. 6) "WORKING" message flashes at bottom right of screen during the recognition process. 7) While "WORKING" is displayed, hitting the space bar will cause the recognition on the current line to stop and the OCR will display what is has recognised thus far and then move on to the next line of text. 8) Only the Escape key will now abandon a run completely. 9) A list of all existing .TIF files will be shown before you enter your TIF file name. 10) The text file name will default to the same name as your TIF file and you are also shown all your .TXT files before you enter your TXT file name. 11) You can choose file names either by using the cursor keys as described below, OR by typing the number shown in front of the filename and then pressing the <Enter> key. List of files ============= OCR.EXE the OCR program for GS4000/GS2000 REMOVE.EXE a utility to remove bad characters from a font-file README.DOC this file COURIER.OCR a fontfile ( all fontfiles must have the extension .OCR ) HELVETIC.OCR a fontfile for Helvetica fonts. TIMROMAN.OCR a fontfile for Times Roman fonts. Installing the software ======================= The software can be run from floppy, hard or RAM discs. To install on a hard disc C: A:>C: C:>MD \OCR C:>CD \OCR C:>COPY A:\*.* /V ( You can install OCR in a different directory, or on a floppy disc, as necessary. The configuration file should be in the current working directory when you run the software. ) Running the software ==================== The OCR program can be run in two modes, either by supplying all options on the command line, or via a menu screen, (however you can not use the Merge facility from the command line). (a) Menu screen --------------- At the DOS prompt, type 'OCR' to use a GS4000 or 'OCR -2' to use a GS2000 and press enter, e.g. C:>OCR -2 The program will then display a menu screen, thus: ----------------------------------------------------------------------------- Resolution (DPI) GENIUS GS2000 200 PRODIGY OCR V1.69 F1: Merge Scan (ON / OFF) __OFF F6: Learn mode ( ON/OFF) OFF F2: Character spacing F7: Scanner /TIF file SCANNER F3: Automatic detection F8: Select font (.OCR) COURIER F4: Sensitivity (-9 to +9) 0 F9: Change Font file path F5: Touching Characters ___ON F10: Change image file path <Enter> Start OCR <esc> Exit to DOS ----------------------------------------------------------------------------- ( File selection area ) -------------(c) 1988,1989 Synergy (UK) Ltd. All rights reserved. ----------- The main box shows which keys to use to control the software, and the lower box is used to display lists of filenames for selection. Pressing F1 allows you to scan text up to 210mm (8 inches wide). For scanning text wider than 105mm you can do 2 scans a left and a right scan, and then merge the text together. This merge facility is enable by pressing F2 to the ON option. To disable this feature, press F2 again to get the OFF option. NOTE: If the merge option is on the next two scans will ALWAYS be merged. SEE BELOW FOR DETAILED DISCUSSION OF THE MERGE FACILITY. Pressing F2 you can force the OCR software to treat the text to be scanned as Mono Spaced(Fixed pitch), Proportionally spaced, or let the software work it out for itself (Automatic detection). If you select fixed pitch then you must enter the pitch size in characters per inch using F3. If you select proportional then you must use F3 to enter a word gap in pixels. (Typical values can be 4,5 at 200dpi, 6,7 at 300dpi, 8-10 at 400dpi). This will help eliminate any extra spaces between letters that form one word when the system uses its automatic mode. F3 has no effect if you select Automatic detection. F4 allows you to adjust the sensitivity of the recognition algorithm. Normally the software uses a sensitivity value of 0 in normal mode and +6 in learn mode. If you wish to make the software more critical then increase the above values, if you want to make the software less critical then decrease the above values. F5 allows you to manually control the touching character algorithm. This will help recognise characters that are touching. However, this algorithm works best with a GOOD dictionary. If you are in learn mode you will find it ususally better to turn this OFF since you won't have a good enough font for it to work well. The default is ON in free run and OFF in learn mode, however you can override these defaults by using F5 AFTER you have used F6 to select your desired mode. Pressing F6 toggles Learn-Mode on/off. When it is off ( the normal state ), characters that are not recognised are replaced by another character ( by default '@' ) and the process proceeds automatically. If learn-mode is on, characters which are unrecognised, or recognisable but poorly formed, are displayed on the screen enlarged. The user may enter what the character is, or skip over it. Pressing F7 toggles between the scanner and image files. You can either perform recognition as you scan the page in, or select TIF and recognise text from a previously scanned .TIF file. The F8 key allows you to select a different font from the one displayed. Press F8 and a list a font files is displayed. Choose one with the cursor keys and press Enter to select it, or Esc. to go back to the original menu. If the font you want is not in the window, press PgUp/PgDn until it appears. You can also select the file by typing the number that preceeds the filename. Key F9 is used to change the default path for the .OCR files. Press F9 and the program shows the current path, and prompts you to enter a new path. To keep the old path just press <Enter> ( or <Esc> ), otherwise type in a new path. Key F10 is used to change the default path for the .TIF files, and for saving TIF files produced when scanning with the scanner. Press F10 and the program shows the current path, and prompts you to enter a new path. To keep the old path just press <Enter> ( or <Esc> ), otherwise type in a new path. Once the parameters are selected, pressing the Enter key will start the recognition process. Alternatively, pressing Esc. will write your configuration to OCR.CFG and exit to DOS. Once Enter is pressed the program will either start up the scanner ( if SCANNER selected ), or ask you to choose an image file ( if TIF selected ). If the SCANNER is selected then you will be prompted for the name of a file in which to save the image that the scanner produces. Note that this image file will be put in the directory that you specified with F10 (above). The image filename is selected in the same way as a font file, with the cursor and PgUp/PgDn keys. Press enter to select the file, or Esc. to go back to the menu. Whether SCANNER or TIF is selected, you must now type in a name for the output text file, e.g. textfile.txt or b:\test.doc Pressing Esc instead will return you to the menu. NOTE -- If you want your output to go into the same file as the previous ==== scan, then just press return to use the filename as shown on the line below the the prompt. If you want the output to go to a new file then you MUST change the filename. Once the recognition process has finished, the program returns you to the menu screen, (UNLESS Merge Scan is ON see below). (b) DOS command line -------------------- You can bypass the menu screen by supplying all the parameters on the DOS command line. The parameters are listed below, and are in the same format as the OCR.CFG file. Parameters may be preceded by '-' or '/'. The only mandatory parameter is a filename for the output. e.g. OCR test.txt -s -s selects the scanner. OCR abc.doc /Imanual.img /I,/i,-I,-i all select imagefile input OCR /h /h or -h lists the options (c) Learn mode -------------- When you run the program in learn mode, the screen is split into two parts. The top part shows the scanned text input as pixels, and the bottom part is used for teaching or examining characters. Characters are shown with a cursor around the pixels that the software thinks that the character is comprised. In the lower part of the screen, the current text line will be shown next to the '->' symbol, and one of the characters is being offered for verification as shown by the ^. To teach the character just type it and press Enter. You will then see the ^ advance to the next character and the cursor will move to the next characters pixels in the top part of the screen. Just pressing Enter will skip over the character to the next one on the line (if any), <esc> will quit, and F10 will save the current state of the font data file, if you type Enter to the prompt SAVE DATA. TEACHING A NEW FONT =================== NOTE -- If you press F1 or the Del key you will delete the character currently ==== above the ^ from the currently loaded font file. A message will appear to indicate that it has been deleted. Note that the deletion will not affect your file on disk unless you finish the entire page in learn mode, OR you use F10 to dump the font file. You may move along the line of characters shown, by using the cursor keys on the numeric keypad. "Left arrow" & "Right arrow" move you one character to left or right respectively. If you press "Down arrow" the cursor will move to the next rightmost character that has a probability of less than 88%. If there is no such character, the cursor will not move. Pressing "Up arrow" is the same except that the next leftmost character will be found. The Home key will move you to the first character on the line, and the End key will move to the last character on the line. When you enter a new character and press "return", the cursor will move to the next character below 88% if there is one, otherwise it will just move to the next character. If the ^ is under the first character on the line and it has a probability > 88%, then you do not need to edit this line unless you can see an obvious mistake. When you are satisfied with the current line, press F5, and the OCR software will proceed to the next line. NOTE You will also see a display in the lower part of the screen which ==== indicates which characters are in the dictionary and which are not. The fact that a character appears in the list of "Taught Characters" does not mean that you don't NEED to teach it. Your only real guide to this is the probability figure as discussed above. ----------------------------------------------------------------------------- Configuration /Command line options =================================== These are the options stored in the configuration file OCR.CFG. They may also be entered on the command line if you do not want to use the menu screen, e.g. Key F10 is used to change the default path for the .TIF files. Press F10 and the program shows the current path, and prompts you to enter a new path. To keep the old path just press <Enter> ( or <Esc> ), otherwise type in a new path. OCR textfile.txt -ipage1.img -l100 will run the software, produce an output text file 'textfile.txt' from the image file 'page1.img', with a scan width of 100 mm. OCR -h or OCR /h will produce a list of the options. -ma mode -ma learn OFF (auto), -ml learn ON -s scanner is source for image -i<imagefile> .TIF file is source -p. default path for .TIF files ( . = current directory ) -fCOURIER font-file (.OCR) -o. default path for .OCR files ( . = current directory ) -d. default path for text output files ( . = current directory ) -h help - list options -$ units displayed in inches ( default is mm. ). Values shown above are system defaults. The option '-k' will write the command-line options to OCR.CFG if entered on the command-line, e.g. ocr -$ -k will set units to inches, write the configuration file, and exit. This method can be used to change any of the options not available on the main menu. Alternatively, OCR.CFG may be edited with a simple ASCII editor, or a word- processor in 'text' or 'ascii' mode. The Configuration File ====================== The configuration file is a plain ASCII text file, which can be edited with EDLIN or most word-processors. A typical file might look like this: OCR.CFG ------- -pC:\IMAGES -d. -o. -fCOURIER -ml -iTEST.TIF -$ *** NB *** The only way to reset to millimetres from inches is to delete the line "-$" from the .cfg file using an editor. Creating a new font file ======================== A new font data file can be created either by modifying an existing file, or by making one from scratch. The new file must have the extension OCR. To add new characters to an existing font, copy its .OCR file to a new file with the same extension, select this as your font file when you run OCR, and use learn mode to teach it the new characters, e.g. C:>cd \ocr C:>copy courier.ocr courier2.ocr /v C:>ocr and select font courier2.ocr and learn-mode ON. To create a completely new font, you must start with a 'blank' fontfile, and teach it characters with learn mode. Run OCR, and press F8 to select a font. Choose the 'Create font' option. The program now asks you for a filename, makes an empty fontfile for you, and selects it as the current font. When teaching a new font, it is a good idea to scan in a sheet of all the characters, something like this: ABCDEFGHIJKLMNOPQRSTUVWXYZ 0123456789 -=# [] {} -=#,.;: abcdefghijklmnopqrstuvwxyz !"£$%^&*() \| /? '@ <> ABCDEFGHIJKLMNOPQRSTUVWXYZ 0123456789 -=# [] {} -=#,.;: abcdefghijklmnopqrstuvwxyz !"£$%^&*() \| /? '@ <> ABCDEFGHIJKLMNOPQRSTUVWXYZ 0123456789 -=# [] {} -=#,.;: abcdefghijklmnopqrstuvwxyz !"£$%^&*() \| /? '@ <> ... and so on. When starting with a brand new dictionary, you may find that a number of the lines at the top of the page being scanned are ignored or produce "noise" characters. Just persevere, the system will soon work out all it needs to scan properly. Scanning text wider than 105mm (4 in). ====================================== If you select the Merge Scan option using the F1 key, the OCR will work as described above, except that:- 1) The text file name that you use will be the name of the MERGED text. 2) You must first scan the left hand section of the text. This will then be OCR'd as usual and when you accept it, it gets saved in a file LEFT.TXT. The scanner light is then enabled and you can scan the right hand section of the text. YOU MUST ENSURE THAT THERE IS AT LEAST 12mm (0.5in) OVERLAP BETWEEN THE TWO SCANS. Don't worry the the left hand text lines end with funny characters and that the right hand text line start with funny characters. This is because you can not always get a whole number of characters in the scan window. 3) After you accept the second scan, the Merge facility will run with the following screen display. -----------left.txt-----------------|-----------------Right.txt--------------- | | | | | | left | right | hand | hand | text | text | | | | | | | ------------------------------------|----------------------------------------- <F10> Save, <Esc> Exit, <F5> Merge, <F6> Toggle display, <cursor keys> Align To use the merge facility you must identify any two halves of a full line in each window that can be merged. To do this you can use the left & right arrow keys to swap between the two windows above, and the up & down keys and the PgUp & PgDn keys to scroll the windows. You will notice that the line in the centre of each window is highlighted. This is the line that must be matched for the merge. When you have two suitable halves highlighted simply press the <F5> key to do the merge. The screen will then swap to display the merged text. If the merge is not good enough, you can press <F6> and try two different lines for the merge. To exit from merge use <Esc>. You can save the merged text at any time by pressing <F10>. The Remove Utility ================== The OCR package comes with a utility, REMOVE.EXE, which allows you to remove from a font-file bad or wrongly entered characters. For example, you may have mistakenly selected 'C' in learn mode instead of 'V'. The only way to correct this is to remove all 'C's in the font, and re-teach them. REMOVE cannot selectively delete only certain instances of a character, all instances will be removed. The simplest way to use the utility is to go to your OCR directory and type REMOVE The program will then prompt you for the name of the font-file ( minus the '.OCR' extension ), and the character(s) you wish to delete. A backup copy of your font-file will be made, with the same name and the extension '.$CR', e.g. C:\OCR\REMOVE REMOVE - Utility to remove characters from a font file. (c) 1988 Synergy (UK) Ltd. All rights reserved. font-file : COURIER characters : RP Reading file :COURIER.OCR Backup file :COURIER.$CR Writing file :COURIER.OCR Deleting characters :RP No. of deletions : 214 In the above example, all occurrences of the characters 'R' and 'P' are removed from COURIER.OCR, and the old version of COURIER.OCR is copied to a file called COURIER.$CR. The full format of the command line is REMOVE [ < font-file> [ < character-list > ] ] so you could also type REMOVE COURIER and only be prompted for the character(s) to be removed, or REMOVE COURIER a?"W which will remove the four characters 'a', '?', '"', and 'W' from COURIER.OCR without further prompting. If the program reports the number of deletions to be zero, then no occurrences of any of the characters you listed were found. Notes ===== 1.) If a particular page gives poor results, try adjusting the contrast - if the print is light increase the contrast, otherwise decrease it if the print is dark. Obviously this only applies when working directly from the scanner. 2.) When teaching a new font the FULL HEIGHT indicator should be ignored on the first few lines of text. It will eventually correct itself and you do not lose the information you appear to put in incorrectly. 3.) DOS 2.XX users -------------- The default path for both .OCR files and text output files is '.', which is the DOS abbreviation for the current working directory. If your ver- sion of DOS does not recognize this abbreviation, you must specify an explicit path either by editing the OCR.CFG file or setting these opt- ions from the command line, e.g. Floppy-disc example: Hard-disc example: edit OCR.CFG to read -oA:\ -oC:\OCR ( to change the .OCR path ) -dA:\ -dC:\OCR ( to change the text file path ) or go to the directory containing OCR.CFG, and type OCR -oA:\ -dA:\ -k OCR -oC:\OCR -dC:\OCR -k ( The '-k ' option causes the program to save the preceding options in OCR.CFG ). 4.) REMOVE.EXE - When entering the double-quote character on the command line, you must precede it with '\', e.g. REMOVE COURIER s\"q will remove the three characters 's', '"', and 'q'. This is necessary because the parameters for a DOS command line may optionally be enclosed in quotes, thus REMOVE COURIER "sq" will remove only 's' and 'q'. This does NOT apply when entering characters in response to the programs own prompt, e.g. C:>REMOVE REMOVE - Utility to remove characters from a font file. (c) 1988 Synergy (UK) Ltd. All rights reserved. font-file : courier characters : s\"q Reading file :courier.OCR Backup file :courier.$CR Writing file :courier.OCR Deleting characters :s\"q No. of deletions : 81 will remove all occurrences of 's', '\', '"', and 'q'. 5.) Colour Graphics Adapter ( CGA ) ------------------------------- If your are using the IBM CGA or a compatible graphics card, you will need to load the Graphic Character Set. ( This is not necessary when using EGA or VGA ). If the graphic characters are not loaded some of the line drawing used on the screen will not be displayed correctly. This affects only the appearance of the screen, the OCR software will perform normally in all other ways. To load the graphic characters, go to the directory containing your DOS files and type GRAFTABL