// Computes the noise_density_ by summing the number of elements in a // neighbourhood of each grid cell. void StrokeWidth::ComputeNoiseDensity(TO_BLOCK* block, TabFind* line_grid) { // Run a preliminary strokewidth neighbour detection on the medium blobs. line_grid->InsertBlobList(true, true, false, &block->blobs, false, this); BLOBNBOX_IT blob_it(&block->blobs); for (blob_it.mark_cycle_pt(); !blob_it.cycled_list(); blob_it.forward()) { SetNeighbours(false, blob_it.data()); } // Remove blobs with a good strokewidth neighbour from the grid. for (blob_it.mark_cycle_pt(); !blob_it.cycled_list(); blob_it.forward()) { BLOBNBOX* blob = blob_it.data(); if (blob->GoodTextBlob() > 0) RemoveBBox(blob); blob->ClearNeighbours(); } // Insert the smaller blobs into the grid. line_grid->InsertBlobList(true, true, false, &block->small_blobs, false, this); line_grid->InsertBlobList(true, true, false, &block->noise_blobs, false, this); if (noise_density_ != NULL) delete noise_density_; IntGrid* cell_counts = CountCellElements(); noise_density_ = cell_counts->NeighbourhoodSum(); delete cell_counts; // Clear the grid as we don't want the small stuff hanging around in it. Clear(); }
// Creates and returns a Pix with the same resolution as the original // in which 1 (black) pixels represent likely non text (photo, line drawing) // areas of the page, deleting from the blob_block the blobs that were // determined to be non-text. // The photo_map is used to bias the decision towards non-text, rather than // supplying definite decision. // The blob_block is the usual result of connected component analysis, // holding the detected blobs. // The returned Pix should be PixDestroyed after use. Pix* CCNonTextDetect::ComputeNonTextMask(bool debug, Pix* photo_map, TO_BLOCK* blob_block) { // Insert the smallest blobs into the grid. InsertBlobList(&blob_block->small_blobs); InsertBlobList(&blob_block->noise_blobs); // Add the medium blobs that don't have a good strokewidth neighbour. // Those that do go into good_grid as an antidote to spreading beyond the // real reaches of a noise region. BlobGrid good_grid(gridsize(), bleft(), tright()); BLOBNBOX_IT blob_it(&blob_block->blobs); for (blob_it.mark_cycle_pt(); !blob_it.cycled_list(); blob_it.forward()) { BLOBNBOX* blob = blob_it.data(); double perimeter_area_ratio = blob->cblob()->perimeter() / 4.0; perimeter_area_ratio *= perimeter_area_ratio / blob->enclosed_area(); if (blob->GoodTextBlob() == 0 || perimeter_area_ratio < kMinGoodTextPARatio) InsertBBox(true, true, blob); else good_grid.InsertBBox(true, true, blob); } noise_density_ = ComputeNoiseDensity(debug, photo_map, &good_grid); good_grid.Clear(); // Not needed any more. Pix* pix = noise_density_->ThresholdToPix(max_noise_count_); if (debug) { pixWrite("junknoisemask.png", pix, IFF_PNG); } ScrollView* win = NULL; #ifndef GRAPHICS_DISABLED if (debug) { win = MakeWindow(0, 400, "Photo Mask Blobs"); } #endif // GRAPHICS_DISABLED // Large and medium blobs are not text if they overlap with "a lot" of small // blobs. MarkAndDeleteNonTextBlobs(&blob_block->large_blobs, kMaxLargeOverlapsWithSmall, win, ScrollView::DARK_GREEN, pix); MarkAndDeleteNonTextBlobs(&blob_block->blobs, kMaxMediumOverlapsWithSmall, win, ScrollView::WHITE, pix); // Clear the grid of small blobs and insert the medium blobs. Clear(); InsertBlobList(&blob_block->blobs); MarkAndDeleteNonTextBlobs(&blob_block->large_blobs, kMaxLargeOverlapsWithMedium, win, ScrollView::DARK_GREEN, pix); // Clear again before we start deleting the blobs in the grid. Clear(); MarkAndDeleteNonTextBlobs(&blob_block->noise_blobs, -1, win, ScrollView::CORAL, pix); MarkAndDeleteNonTextBlobs(&blob_block->small_blobs, -1, win, ScrollView::GOLDENROD, pix); MarkAndDeleteNonTextBlobs(&blob_block->blobs, -1, win, ScrollView::WHITE, pix); if (debug) { #ifndef GRAPHICS_DISABLED win->Update(); #endif // GRAPHICS_DISABLED pixWrite("junkccphotomask.png", pix, IFF_PNG); #ifndef GRAPHICS_DISABLED delete win->AwaitEvent(SVET_DESTROY); delete win; #endif // GRAPHICS_DISABLED } return pix; }