Improved results on images with no resolution. Estimates resolution

from the size of the connected components, based on average text size.

Author: Ray Smith

api/baseapi.cpp

@@ -102,15 +102,6 @@ const char* kInputFile = "noname.tif";
 const char* kOldVarsFile = "failed_vars.txt";
 /** Max string length of an int.  */
 const int kMaxIntSize = 22;
-/**
- * Minimum believable resolution. Used as a default if there is no other
- * information, as it is safer to under-estimate than over-estimate.
- */
-const int kMinCredibleResolution = 70;
-/** Maximum believable resolution.  */
-const int kMaxCredibleResolution = 2400;
-/** Default resolution.  */
-const int kDefaultResolution = 300;
 
 /* Add all available languages recursively.
 */
@@ -2225,8 +2216,8 @@ bool TessBaseAPI::Threshold(Pix** pix) {
     // Use the minimum default resolution, as it is safer to under-estimate
     // than over-estimate resolution.
     tprintf("Warning. Invalid resolution %d dpi. Using %d instead.\n", y_res,
-            kDefaultResolution);
-    thresholder_->SetSourceYResolution(kDefaultResolution);
+            kMinCredibleResolution);
+    thresholder_->SetSourceYResolution(kMinCredibleResolution);
   }
   PageSegMode pageseg_mode =
       static_cast<PageSegMode>(

ccmain/osdetect.cpp

@@ -58,9 +58,6 @@ const char* ScriptDetector::korean_script_ = "Korean";
 const char* ScriptDetector::japanese_script_ = "Japanese";
 const char* ScriptDetector::fraktur_script_ = "Fraktur";
 
-// Minimum believable resolution.
-const int kMinCredibleResolution = 70;
-
 void OSResults::update_best_orientation() {
   float first = orientations[0];
   float second = orientations[1];

ccmain/pagesegmain.cpp

@@ -310,13 +310,22 @@ ColumnFinder* Tesseract::SetupPageSegAndDetectOrientation(
   TO_BLOCK* to_block = to_block_it.data();
   TBOX blkbox = to_block->block->bounding_box();
   ColumnFinder* finder = NULL;
+  int estimated_resolution = source_resolution_;
+  if (source_resolution_ == kMinCredibleResolution) {
+    // Try to estimate resolution from typical body text size.
+    int res = IntCastRounded(to_block->line_size * kResolutionEstimationFactor);
+    if (res > estimated_resolution && res < kMaxCredibleResolution) {
+      estimated_resolution = res;
+      tprintf("Estimating resolution as %d\n", estimated_resolution);
+    }
+  }
 
   if (to_block->line_size >= 2) {
     finder = new ColumnFinder(static_cast<int>(to_block->line_size),
                               blkbox.botleft(), blkbox.topright(),
-                              source_resolution_, textord_use_cjk_fp_model,
-                              textord_tabfind_aligned_gap_fraction,
-                              &v_lines, &h_lines, vertical_x, vertical_y);
+                              estimated_resolution, textord_use_cjk_fp_model,
+                              textord_tabfind_aligned_gap_fraction, &v_lines,
+                              &h_lines, vertical_x, vertical_y);
 
     finder->SetupAndFilterNoise(pageseg_mode, *photo_mask_pix, to_block);
 

ccstruct/publictypes.h

@@ -30,7 +30,19 @@
 // API-level code should include apitypes.h in preference to this file.
 
 /** Number of printers' points in an inch. The unit of the pointsize return. */
-const int kPointsPerInch = 72;
+constexpr int kPointsPerInch = 72;
+/**
+ * Minimum believable resolution. Used as a default if there is no other
+ * information, as it is safer to under-estimate than over-estimate.
+ */
+constexpr int kMinCredibleResolution = 70;
+/** Maximum believable resolution.  */
+constexpr int kMaxCredibleResolution = 2400;
+/**
+ * Ratio between median blob size and likely resolution. Used to estimate
+ * resolution when none is provided. This is basically 1/usual text size in
+ * inches.  */
+constexpr int kResolutionEstimationFactor = 10;
 
 /**
  * Possible types for a POLY_BLOCK or ColPartition.