/** * Auto page segmentation. Divide the page image into blocks of uniform * text linespacing and images. * * Resolution (in ppi) is derived from the input image. * * The output goes in the blocks list with corresponding TO_BLOCKs in the * to_blocks list. * * If single_column is true, then no attempt is made to divide the image * into columns, but multiple blocks are still made if the text is of * non-uniform linespacing. * * If osd (orientation and script detection) is true then that is performed * as well. If only_osd is true, then only orientation and script detection is * performed. If osd is desired, (osd or only_osd) then osr_tess must be * another Tesseract that was initialized especially for osd, and the results * will be output into osr (orientation and script result). */ int Tesseract::AutoPageSeg(PageSegMode pageseg_mode, BLOCK_LIST* blocks, TO_BLOCK_LIST* to_blocks, Tesseract* osd_tess, OSResults* osr) { if (textord_debug_images) { WriteDebugBackgroundImage(textord_debug_printable, pix_binary_); } Pix* photomask_pix = NULL; Pix* musicmask_pix = NULL; // The blocks made by the ColumnFinder. Moved to blocks before return. BLOCK_LIST found_blocks; TO_BLOCK_LIST temp_blocks; bool single_column = !PSM_COL_FIND_ENABLED(pageseg_mode); bool osd_enabled = PSM_OSD_ENABLED(pageseg_mode); bool osd_only = pageseg_mode == PSM_OSD_ONLY; ColumnFinder* finder = SetupPageSegAndDetectOrientation( single_column, osd_enabled, osd_only, blocks, osd_tess, osr, &temp_blocks, &photomask_pix, &musicmask_pix); int result = 0; if (finder != NULL) { TO_BLOCK_IT to_block_it(&temp_blocks); TO_BLOCK* to_block = to_block_it.data(); if (musicmask_pix != NULL) { // TODO(rays) pass the musicmask_pix into FindBlocks and mark music // blocks separately. For now combine with photomask_pix. pixOr(photomask_pix, photomask_pix, musicmask_pix); } if (equ_detect_) { finder->SetEquationDetect(equ_detect_); } result = finder->FindBlocks(pageseg_mode, scaled_color_, scaled_factor_, to_block, photomask_pix, pix_thresholds_, pix_grey_, &found_blocks, to_blocks); if (result >= 0) finder->GetDeskewVectors(&deskew_, &reskew_); delete finder; } pixDestroy(&photomask_pix); pixDestroy(&musicmask_pix); if (result < 0) return result; blocks->clear(); BLOCK_IT block_it(blocks); // Move the found blocks to the input/output blocks. block_it.add_list_after(&found_blocks); if (textord_debug_images) { // The debug image is no longer needed so delete it. unlink(AlignedBlob::textord_debug_pix().string()); } return result; }
/** * Auto page segmentation. Divide the page image into blocks of uniform * text linespacing and images. * * Resolution (in ppi) is derived from the input image. * * The output goes in the blocks list with corresponding TO_BLOCKs in the * to_blocks list. * * If !PSM_COL_FIND_ENABLED(pageseg_mode), then no attempt is made to divide * the image into columns, but multiple blocks are still made if the text is * of non-uniform linespacing. * * If diacritic_blobs is non-null, then diacritics/noise blobs, that would * confuse layout anaylsis by causing textline overlap, are placed there, * with the expectation that they will be reassigned to words later and * noise/diacriticness determined via classification. * * If osd (orientation and script detection) is true then that is performed * as well. If only_osd is true, then only orientation and script detection is * performed. If osd is desired, (osd or only_osd) then osr_tess must be * another Tesseract that was initialized especially for osd, and the results * will be output into osr (orientation and script result). */ int Tesseract::AutoPageSeg(PageSegMode pageseg_mode, BLOCK_LIST* blocks, TO_BLOCK_LIST* to_blocks, BLOBNBOX_LIST* diacritic_blobs, Tesseract* osd_tess, OSResults* osr) { Pix* photomask_pix = NULL; Pix* musicmask_pix = NULL; // The blocks made by the ColumnFinder. Moved to blocks before return. BLOCK_LIST found_blocks; TO_BLOCK_LIST temp_blocks; ColumnFinder* finder = SetupPageSegAndDetectOrientation( pageseg_mode, blocks, osd_tess, osr, &temp_blocks, &photomask_pix, &musicmask_pix); int result = 0; if (finder != NULL) { TO_BLOCK_IT to_block_it(&temp_blocks); TO_BLOCK* to_block = to_block_it.data(); if (musicmask_pix != NULL) { // TODO(rays) pass the musicmask_pix into FindBlocks and mark music // blocks separately. For now combine with photomask_pix. pixOr(photomask_pix, photomask_pix, musicmask_pix); } if (equ_detect_) { finder->SetEquationDetect(equ_detect_); } result = finder->FindBlocks(pageseg_mode, scaled_color_, scaled_factor_, to_block, photomask_pix, pix_thresholds_, pix_grey_, &pixa_debug_, &found_blocks, diacritic_blobs, to_blocks); if (result >= 0) finder->GetDeskewVectors(&deskew_, &reskew_); delete finder; } pixDestroy(&photomask_pix); pixDestroy(&musicmask_pix); if (result < 0) return result; blocks->clear(); BLOCK_IT block_it(blocks); // Move the found blocks to the input/output blocks. block_it.add_list_after(&found_blocks); return result; }