In This Topic ▼

L_AutoZoneBitmap

Summary

Automatically detects different zone types (Text, Graphics and Tables) in an image. Classifying by types, improves the recognition results from OCR pre-processing. This function is useful for any application that needs to automatically separate images, tables and text within mixed raster content (MRC) images.

Syntax

#include "l_bitmap.h"

L_LTIMGCOR_API L_INT EXT_FUNCTION L_AutoZoneBitmap(pBitmap, phZones, puCount, uFlags, pCallback, pUserData)

Parameters

pBITMAPHANDLE pBitmap

Pointer to the bitmap handle referencing the bitmap in which the zones will be detected.

pLEADZONE * phZones

Pointer to a handle to be updated with the detected zones. To retrieve data about the detected zones cast the phZones to pLEADZONE and call the GlobalLock function.

L_UINT32 * puCount

Pointer to a variable to be updated with the number of detected zones.

L_UINT32 uFlags

Flags indicating how the function should behave. Combine values when appropriate by using a bitwise OR ( | ). Possible values are:

Value	Meaning
Types of zones to detect (At least one is required):
AUTOZONE_DETECT_TEXT	[0x0001] Detect text zones.
AUTOZONE_DETECT_GRAPHIC	[0x0002] Detect graphic zones.
AUTOZONE_DETECT_TABLE	[0x0004] Detect table zones.
AUTOZONE_DETECT_ALL	[0x00010007] Detect all zone types (Text, Graphics and Tables).
Allow overlapping among zones (Optional):
AUTOZONE_DONT_ALLOW_OVERLAP	[0x0000] Do not allow zones to overlap.
AUTOZONE_ALLOW_OVERLAP	[0x0010] Allow zones to overlap.
How to merge text zones (Optional):
AUTOZONE_ACCURATE_ZONES	[0x0000] Do not merge text zones and keep them separated (paragraphs).
AUTOZONE_GENERAL_ZONES	[0x0100] Merge text zones as much as possible.
Table zone detection options (Optional):
AUTOZONE_DONT_RECOGNIZE_ONE_CELL_TABLE	[0x0000] Ignore one-cell tables (borders), and detect what is inside.
AUTOZONE_RECOGNIZE_ONE_CELL_TABLE	[0x1000] Consider borders as one-cell tables.
AUTOZONE_NORMAL_TABLE (version 17 or later)	[0x0000] Use Normal Table Detection.
AUTOZONE_ADVANCED_TABLE (version 17 or later)	[0x2000] Use Advanced Table Detection which returns more accurate results and detects complex tables.
AUTOZONE_LINES_RECONSTRUCTION (version 17 or later)	[0x4000] Use Line Reconstruction for patterned tables and to connect broken lines.
Multi-threading usage (Optional):
AUTOZONE_USE_MULTITHREADING	[0x00000000] Use multi-threading (faster for multi-core CPUs).
AUTOZONE_DONTUSE_MULTITHREADING	[0x80000000] Do not use multi-threading (necessary for single-core CPUs).
Bitmap should be changed so all it contains is text (Optional):
AUTOZONE_TEXT_DETECTION	[0x8000] If set, the function does not return text zones in `phZones`. Instead, it modifies `pBitmap` to have text areas only: graphics and tables areas are deleted from the input image. NOTE: Make a copy of the original image if you want to keep it. When this function is set to AUTOZONE_TEXT_DETECTION, the original image is modified.
Type of the document needed to be autozoned (Optional):
AUTOZONE_TEXTBOOK	[0x40000] Setting this flag improves zoning book's pages.
AUTOZONE_FAVOR_GRAPHICS	[0x400000] This flag is used to improve detecting figure zones in documents.
Type of the document needed to be auto-zoned (Optional):
AUTOZONE_DETECT_VERTICAL_TEXT	[0x100000] Set this flag if document image contains vertical text lines.
Detect checkbox zones and the sensitivity of the detection (Optional):
AUTOZONE_DETECT_CHECKBOX	[0x10000] Set this flag to detect checkbox zones in documents.
AUTOZONE_CHECKBOX_SENSITIVITY_HIGH	[0x00000000] The sensitivity of matching the checkboxes shape is high, and hence, the false negative detections is low.
AUTOZONE_CHECKBOX_SENSITIVITY_LOW	[0x200000] The sensitivity of matching the checkboxes shape is low, and hence, the false negative detections is high.
Document contains Asian zones:
AUTOZONE_ASIAN_ZONING	[0x200] Detect Asian text (Japanese, Chinese, Korean, etc.)
Document contains handwritten zones
AUTOZONE_ICR_ZONING	[0x00000080] Detect handwritten zones

AUTOZONECALLBACK pCallback

Optional callback function for additional processing.

If you do not provide a callback function, use NULL as the value of this parameter.
If you do provide a callback function, use the function pointer as the value of this parameter. The callback function must adhere to the function prototype described in the AUTOZONECALLBACK Function.

L_VOID * pUserData

Void pointer that you can use to pass one or more additional parameters that the callback function needs.

To use this feature, assign a value to a variable or create a structure that contains as many fields as you need. Then, in this parameter, pass the address of the variable or structure, casting it to L_VOID *. The callback function, which receives the address in its own pUserData parameter, can cast it to a pointer of the appropriate data type to access your variable or structure.

If additional parameters are not needed, pass NULL.

Returns

Value	Meaning
SUCCESS	The function was successful.
< 1	An error occurred. Refer to Return Codes.

Comments

After using this function free phZones by using the L_FreeZoneData function.

To update a status bar or detect a user interrupt during execution of this function, see the L_SetStatusCallback function.

This function does not support 12 and 16-bit grayscale and 48 and 64-bit color images. If the image is 12 and 16-bit grayscale and 48 and 64-bit color, the function will not return an error.

This function does not support signed data images. It returns the error code ERROR_SIGNED_DATA_NOT_SUPPORTED if a signed data image is passed to this function.

This function does not support 32-bit grayscale images. It returns the ERROR_GRAY32_UNSUPPORTED error code if a 32-bit grayscale image is passed to this function.

Required DLLs and Libraries

LTIMGCOR
For a listing of the exact DLLs and Libraries needed, based on the toolkit version, refer to Files To Be Included With Your Application.

Platforms

Win32, x64, Linux.

Example

Detects the image zones and return the results.

L_INT AutoZoneBitmapExample(L_VOID) 
{ 
   L_INT nRet; 
   BITMAPHANDLE LeadBitmap;   /* Bitmap handle to hold the loaded image. */ 
   pLEADZONE pZones = NULL; 
   L_UINT32 count = 0; 
   L_INT TextZonesCount = 0; 
   L_INT GraphicZonesCount = 0; 
   L_INT TableZonesCount = 0; 
   L_INT CellsCount = 0; 
   L_UINT i; 
 
   nRet = L_LoadBitmap(MAKE_IMAGE_PATH(TEXT("Clean.tif")), &LeadBitmap, sizeof(BITMAPHANDLE), 0, ORDER_BGR, NULL, NULL); 
   if (nRet != SUCCESS) 
      return nRet; 
 
 
   nRet = L_AutoZoneBitmap(&LeadBitmap, &pZones, &count, AUTOZONE_DETECT_TEXT, NULL, NULL); 
   if (nRet != SUCCESS) 
      return nRet; 
 
   for (i = 0; i < count; i++) 
   { 
      if (pZones[i].uZoneType == LEAD_ZONE_TYPE_TEXT) 
      { 
         TextZonesCount++; 
      } 
 
      if (pZones[i].uZoneType == LEAD_ZONE_TYPE_GRAPHIC) 
      { 
         GraphicZonesCount++; 
      } 
 
      if (pZones[i].uZoneType == LEAD_ZONE_TYPE_TABLE) 
      { 
         TableZonesCount++; 
         TABLEZONE* pTable = (TABLEZONE*)GlobalLock(pZones[i].pZoneData); 
         CellsCount = pTable->Rows * pTable->Columns; 
         GlobalUnlock(pTable); 
      } 
   } 
 
   nRet = L_FreeZoneData(pZones, count); 
 
   if (nRet != SUCCESS) 
      return nRet; 
 
   if (LeadBitmap.Flags.Allocated) 
      L_FreeBitmap(&LeadBitmap); 
 
   return SUCCESS; 
}