Updated mupdf to improve line detection; added WIP table detection

js/extractTables.js

@@ -1,5 +1,5 @@
-import { calcBoxOverlap } from './modifyOCR.js';
 import ocr from './objects/ocrObjects.js';
+import { calcBoxOverlap } from './utils/miscUtils.js';
 
 /**
  *

js/import/convertPageHocr.js

@@ -7,6 +7,7 @@ import {
 
 import { LayoutDataTablePage } from '../objects/layoutObjects.js';
 import { pass2, pass3 } from './convertPageShared.js';
+import { detectTablesInPage, makeTableFromBbox } from '../utils/detectTables.js';
 
 // If enabled, raw strings are saved in OCR objects for debugging purposes.
 const debugMode = true;
@@ -374,7 +375,18 @@ export async function convertPageHocr({
   pass2(pageObj, rotateAngle);
   const langSet = pass3(pageObj);
 
+  const autoDetectTables = false;
+  const dataTablePage = new LayoutDataTablePage(n);
+  if (autoDetectTables) {
+    const tableBboxes = detectTablesInPage(pageObj);
+    tableBboxes.forEach((bbox) => {
+      const dataTable = makeTableFromBbox(pageObj, bbox);
+      dataTable.page = dataTablePage;
+      dataTablePage.tables.push(dataTable);
+    });
+  }
+
   return {
-    pageObj, dataTables: new LayoutDataTablePage(n), warn, langSet,
+    pageObj, dataTables: dataTablePage, warn, langSet,
   };
 }

js/import/convertPageStext.js

@@ -2,6 +2,7 @@ import ocr from '../objects/ocrObjects.js';
 
 import {
   calcBboxUnion,
+  calcBoxOverlap,
   calcLang,
   mean50,
   quantile,
@@ -10,6 +11,8 @@ import {
 } from '../utils/miscUtils.js';
 
 import { LayoutDataTablePage } from '../objects/layoutObjects.js';
+import { detectTablesInPage, makeTableFromBbox } from '../utils/detectTables.js';
+import { splitLineAgressively } from '../utils/ocrUtils.js';
 
 /**
  * @param {Object} params
@@ -410,6 +413,10 @@ export async function convertPageStext({ ocrStr, n }) {
       // If there are no letters in the line, drop the entire line element
       if (lettersKept === 0) return;
 
+      // Recalculate the bounding box.
+      // The bounding boxes reported by mupdf are often significantly larger than the actual text.
+      ocr.updateLineBbox(lineObj);
+
       pageObj.lines.push(lineObj);
       parLineArr.push(lineObj);
       // eslint-disable-next-line consistent-return
@@ -449,5 +456,32 @@ export async function convertPageStext({ ocrStr, n }) {
 
   pageObj.angle = angleOut;
 
-  return { pageObj, dataTables: new LayoutDataTablePage(n), langSet };
+  const autoDetectTables = false;
+  const dataTablePage = new LayoutDataTablePage(n);
+  if (autoDetectTables) {
+    const tableBboxes = detectTablesInPage(pageObj);
+
+    for (let i = 0; i < pageObj.lines.length; i++) {
+      const line = pageObj.lines[i];
+      let inTable = false;
+      for (let j = 0; j < tableBboxes.length; j++) {
+        if (calcBoxOverlap(line.bbox, tableBboxes[j]) > 0.25) {
+          inTable = true;
+          break;
+        }
+      }
+      if (inTable) {
+        const newLines = splitLineAgressively(line);
+        pageObj.lines.splice(i, 1, ...newLines);
+      }
+    }
+
+    tableBboxes.forEach((bbox) => {
+      const dataTable = makeTableFromBbox(pageObj, bbox);
+      dataTable.page = dataTablePage;
+      dataTablePage.tables.push(dataTable);
+    });
+  }
+
+  return { pageObj, dataTables: dataTablePage, langSet };
 }

js/modifyOCR.js

@@ -1,27 +1,5 @@
 import ocr from './objects/ocrObjects.js';
-import { getRandomAlphanum } from './utils/miscUtils.js';
-
-/**
- * Returns the proportion of boxA's area contained in boxB
- * @param {bbox} boxA
- * @param {bbox} boxB
- */
-export function calcBoxOverlap(boxA, boxB) {
-  const left = Math.max(boxA.left, boxB.left);
-  const top = Math.max(boxA.top, boxB.top);
-  const right = Math.min(boxA.right, boxB.right);
-  const bottom = Math.min(boxA.bottom, boxB.bottom);
-
-  const width = right - left;
-  const height = bottom - top;
-
-  if (width < 0 || height < 0) return 0;
-
-  const areaA = (boxA.bottom - boxA.top) * (boxA.right - boxA.left);
-  const area = width * height;
-
-  return area / areaA;
-}
+import { calcBoxOverlap, getRandomAlphanum } from './utils/miscUtils.js';
 
 /**
  * Adds lines from a new page to an existing page.