@scanmate/ocr 0.2.0 → 0.3.0

package/dist/index.esm.js

@@ -727,8 +727,20 @@ function templateKey(run, character) {
  * @param runs - Its text layer.
  * @returns Glyph images, by {@link templateKey}.
  */
-function collectTemplates(page, dpi, runs) {
-  const templates = new Map();
+/**
+ * Folds one page's glyphs into a store that already holds others.
+ *
+ * Templates are filed by face, size and turn, and a PDF renders those
+ * identically wherever they appear - so a glyph from page 4 is as good a
+ * template as one from page 1, and on many documents it is the only one there
+ * is. Measured on a real order confirmation: the face its page-1 total is set
+ * in carries six distinct digits on that page and nine across the document, and
+ * six is below the bar for checking anything at all.
+ *
+ * The per-character cap applies to the store as a whole, so a long document
+ * costs no more memory than a short one.
+ */
+function collectInto(templates, page, dpi, runs) {
   for (const run of runs) {
     const characters = [...run.text].filter(character => character.trim() !== '');
     const lightOnDark = printPolarity(page, dpi, run) === 'light-on-dark';
@@ -743,6 +755,9 @@ function collectTemplates(page, dpi, runs) {
   }
   return templates;
 }
+function collectTemplates(page, dpi, runs) {
+  return collectInto(new Map(), page, dpi, runs);
+}
 /** Files one glyph under its key, up to the few that are worth keeping. */
 function keep(templates, key, glyph) {
   if (glyph === null) return;
@@ -1457,6 +1472,24 @@ function luminance(data, i) {
 async function ocrPages(pages, options = {}) {
   const engine = options.engine ?? (await createTesseractEngine(options.tesseract));
   try {
+    // The glyphs of the whole document, gathered before any page is read.
+    //
+    // A page's own figures are often set in a face it uses for little else, and
+    // a run cannot be checked against rivals that are not there - on a real
+    // order confirmation the face of the page-1 total carries six distinct
+    // digits on that page and nine across the document, and six is below the
+    // bar for checking anything at all. The same face at the same size renders
+    // identically on every page, so a glyph from page 4 is as good a template
+    // as one from page 1.
+    //
+    // Each page's greyscale is dropped as soon as its glyphs are taken, so this
+    // holds one page of pixels at a time, not the document.
+    const templates = new Map();
+    if (options.printCheck !== false) for (const page of pages) {
+      const items = page.metadata?.original?.textItems;
+      if (items === undefined || items === null || items.length === 0) continue;
+      collectInto(templates, toGrayscale(page.original.raster), page.original.dpi ?? options.assumeDpi ?? 150, items);
+    }
     const results = [];
     for (const [position, page] of pages.entries()) {
       const index = position + 1;
@@ -1468,7 +1501,7 @@ async function ocrPages(pages, options = {}) {
         index,
         total: pages.length
       });
-      const result = await readPage(page, engine, options);
+      const result = await readPage(page, engine, options, templates);
       results.push({
         ...page,
         text: result
@@ -1506,7 +1539,7 @@ async function ocrPages(pages, options = {}) {
     if (options.engine === undefined) await engine.terminate();
   }
 }
-async function readPage(page, engine, options) {
+async function readPage(page, engine, options, templates) {
   const {
     original: originalSource = 'auto',
     targetDpi = 300,
@@ -1600,7 +1633,6 @@ async function readPage(page, engine, options) {
     const printed = items.map(item => ({
       ...item
     }));
-    const templates = collectTemplates(originalGray, originalDpi, printed);
     for (const [r, run] of printed.entries()) {
       const verified = verifyPrintedRun(originalGray, scanGray, originalDpi, run, templates, printCheck);
       if (verified === null) continue;
@@ -1714,5 +1746,5 @@ function mean(values) {
   return values.length === 0 ? 0 : values.reduce((a, b) => a + b, 0) / values.length;
 }
 
-export { DEFAULT_NORMALISE, DEFAULT_RECHECK_PASSES, DEFAULT_TESSERACT_OPTIONS, FIGURE_CHARACTERS$1 as FIGURE_CHARACTERS, TEXT_CHARACTERS, claimWords, collectTemplates, compareTexts, cosine, createTesseractEngine, diacriticsMap, dice, foldConfusables, foldDiacritics, glyphCells, glyphWords, jaccard, jaroWinkler, judgeRun, judgeRuns, levenshtein, levenshteinSimilarity, matchWords, normaliseText, ocrPages, placeGlyphs, printPolarity, readRun, recheckRun, templateKey, tokenise, verifyPrintedRun, wordDistance, wordRecall };
+export { DEFAULT_NORMALISE, DEFAULT_RECHECK_PASSES, DEFAULT_TESSERACT_OPTIONS, FIGURE_CHARACTERS$1 as FIGURE_CHARACTERS, TEXT_CHARACTERS, claimWords, collectInto, collectTemplates, compareTexts, cosine, createTesseractEngine, diacriticsMap, dice, foldConfusables, foldDiacritics, glyphCells, glyphWords, jaccard, jaroWinkler, judgeRun, judgeRuns, levenshtein, levenshteinSimilarity, matchWords, normaliseText, ocrPages, placeGlyphs, printPolarity, readRun, recheckRun, templateKey, tokenise, verifyPrintedRun, wordDistance, wordRecall };
 //# sourceMappingURL=index.esm.js.map

package/dist/src/index.d.ts

@@ -30,8 +30,8 @@ export { compareTexts } from './text-similarity/index.js';
 export type { ScoreMetric, TextMetrics } from './text-similarity/index.js';
 export { DEFAULT_NORMALISE, normaliseText, tokenise } from './text-normalisation/index.js';
 export type { NormaliseOptions } from './text-normalisation/index.js';
-export { collectTemplates, FIGURE_CHARACTERS, glyphCells, glyphWords, placeGlyphs, printPolarity, templateKey, TEXT_CHARACTERS, verifyPrintedRun } from './print-verification/index.js';
-export type { CellOptions, PrintPolarity, PrintVerification, Templates, VerifyOptions } from './print-verification/index.js';
+export { collectInto, collectTemplates, FIGURE_CHARACTERS, glyphCells, glyphWords, placeGlyphs, printPolarity, templateKey, TEXT_CHARACTERS, verifyPrintedRun } from './print-verification/index.js';
+export type { CellOptions, PrintPolarity, PrintVerification, TemplateStore, Templates, VerifyOptions } from './print-verification/index.js';
 export { claimWords, DEFAULT_RECHECK_PASSES, judgeRun, judgeRuns, matchWords, readRun, recheckRun } from './page-reading/index.js';
 export type { Claims, MatchOptions, Recheck, RecheckOptions, RecheckPass, Reference, Verdict, WordMatch } from './page-reading/index.js';
 export { cosine, dice, jaccard, jaroWinkler, levenshtein, levenshteinSimilarity, wordDistance, wordRecall } from './text-similarity/index.js';

package/dist/src/print-verification/glyph-templates.use-case.d.ts

@@ -15,6 +15,8 @@ import type { GrayImage, ScanmateOrientedRect, TextRun } from '@scanmate/ink';
 /** A run of the original's text layer, where it sits and how it is set. */
 /** Glyph images by face, size and character. */
 export type Templates = ReadonlyMap<string, GrayImage[]>;
+/** A store still being filled. Readers take {@link Templates}; only the collector writes. */
+export type TemplateStore = Map<string, GrayImage[]>;
 /** The key a run's glyphs are filed under. */
 export declare function templateKey(run: TextRun, character: string): string;
 /**
@@ -25,6 +27,20 @@ export declare function templateKey(run: TextRun, character: string): string;
  * @param runs - Its text layer.
  * @returns Glyph images, by {@link templateKey}.
  */
+/**
+ * Folds one page's glyphs into a store that already holds others.
+ *
+ * Templates are filed by face, size and turn, and a PDF renders those
+ * identically wherever they appear - so a glyph from page 4 is as good a
+ * template as one from page 1, and on many documents it is the only one there
+ * is. Measured on a real order confirmation: the face its page-1 total is set
+ * in carries six distinct digits on that page and nine across the document, and
+ * six is below the bar for checking anything at all.
+ *
+ * The per-character cap applies to the store as a whole, so a long document
+ * costs no more memory than a short one.
+ */
+export declare function collectInto(templates: TemplateStore, page: GrayImage, dpi: number, runs: readonly TextRun[]): Templates;
 export declare function collectTemplates(page: GrayImage, dpi: number, runs: readonly TextRun[]): Templates;
 /**
  * The greyscale of one box of the page, or `null` when it lies outside it;

package/dist/src/print-verification/index.d.ts

@@ -1,7 +1,7 @@
 export { glyphCells, glyphWords, placeGlyphs } from './glyph-cells.use-case.js';
 export type { CellOptions } from './glyph-cells.use-case.js';
-export { collectTemplates, templateKey } from './glyph-templates.use-case.js';
-export type { Templates } from './glyph-templates.use-case.js';
+export { collectInto, collectTemplates, templateKey } from './glyph-templates.use-case.js';
+export type { TemplateStore, Templates } from './glyph-templates.use-case.js';
 export { mergeVerifiedFigures } from './merge-reading.mapper.js';
 export { printPolarity } from './print-polarity.policy.js';
 export type { PrintPolarity } from './print-polarity.policy.js';

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@scanmate/ocr",
-  "version": "0.2.0",
+  "version": "0.3.0",
   "description": "How closely a scan's text matches the original's: OCR matched run by run on the page, figures held exact, every difference located.",
   "license": "MIT",
   "author": "Eduardo Russo",
@@ -41,7 +41,7 @@
     "!**/*.js.map"
   ],
   "dependencies": {
-    "@scanmate/ink": "^0.2.0",
+    "@scanmate/ink": "^0.3.0",
     "@tesseract.js-data/eng": "^1.0.0",
     "fastest-levenshtein": "^1.0.16",
     "tesseract.js": "^7.0.0"
@@ -50,7 +50,7 @@
     "access": "public"
   },
   "devDependencies": {
-    "@scanmate/align": "^0.2.0",
-    "@scanmate/extract": "^0.2.0"
+    "@scanmate/align": "^0.3.0",
+    "@scanmate/extract": "^0.3.0"
   }
 }