@scanmate/ocr 0.0.2 → 0.0.3

package/dist/index.esm.js

@@ -611,12 +611,12 @@ const DEFAULT_RECHECK_PASSES = [{
 const FIGURE_CHARACTERS = '0123456789.,:/%()+-';
 /** White kept around a crop, in pixels: tesseract reads poorly off the edge of an image. */
 const MARGIN = 12;
-async function recheckRun(engine, image, run, options) {
+async function recheckRun(engine, image, run, options, polarity) {
   const {
     passes = DEFAULT_RECHECK_PASSES,
     agree = 2
   } = options;
-  const crop = cropRun(image.raster, image.dpi, run);
+  const crop = cropRun(image.raster, image.dpi, run, polarity);
   if (crop === null) return {
     cleared: false,
     reading: null,
@@ -632,7 +632,8 @@ async function recheckRun(engine, image, run, options) {
     tried++;
     const scale = Math.max(1, pass.dpi / image.dpi);
     const enlarged = scale === 1 ? crop : await resampleRaster(crop, Math.round(crop.width * scale), Math.round(crop.height * scale));
-    const prepared = frame(pass.stretch === true ? stretch(enlarged) : enlarged);
+    // Inverted text keeps the bar's grey behind it; only a stretch makes that paper white.
+    const prepared = frame(polarity === 'light-on-dark' || pass.stretch === true ? stretch(enlarged) : enlarged);
     const read = await engine.recognise(prepared, {
       layout: pass.layout,
       characters
@@ -658,9 +659,10 @@ async function recheckRun(engine, image, run, options) {
 }
 /**
  * The run's box on the image, grown by a quarter of its height (at least 2 pt)
- * so the glyphs are whole; light text on a dark bar is turned dark on light.
+ * so the glyphs are whole; light text on a dark bar is turned dark on light -
+ * as the original prints it when that is known, else when the crop is dark.
  */
-function cropRun(raster, dpi, run) {
+function cropRun(raster, dpi, run, polarity) {
   const s = dpi / 72;
   const pad = Math.max(2, run.height * 0.25);
   const left = Math.max(0, Math.floor((run.x - pad) * s));
@@ -672,9 +674,29 @@ function cropRun(raster, dpi, run) {
   for (let y = top; y < bottom; y++) out.data.set(raster.data.subarray((y * raster.width + left) * 4, (y * raster.width + right) * 4), (y - top) * out.width * 4);
   let sum = 0;
   for (let i = 0; i < out.data.length; i += 4) sum += luminance(out.data, i);
-  if (sum / (out.width * out.height) < 110) for (let i = 0; i < out.data.length; i += 4) for (let c = 0; c < 3; c++) out.data[i + c] = 255 - out.data[i + c];
+  const invert = polarity === undefined ? sum / (out.width * out.height) < 110 : polarity === 'light-on-dark';
+  if (invert) for (let i = 0; i < out.data.length; i += 4) for (let c = 0; c < 3; c++) out.data[i + c] = 255 - out.data[i + c];
   return out;
 }
+/**
+ * Which way the original prints a run, read off its own crisp rendering: the
+ * glyphs are the pixels far from the background, and the background is most of
+ * the box.
+ */
+function printPolarity(raster, dpi, run) {
+  const s = dpi / 72;
+  const left = Math.max(0, Math.floor(run.x * s));
+  const top = Math.max(0, Math.floor(run.y * s));
+  const right = Math.min(raster.width, Math.ceil((run.x + run.width) * s));
+  const bottom = Math.min(raster.height, Math.ceil((run.y + run.height) * s));
+  const values = [];
+  for (let y = top; y < bottom; y++) for (let x = left; x < right; x++) values.push(luminance(raster.data, (y * raster.width + x) * 4));
+  if (values.length === 0) return 'dark-on-light';
+  const background = values.toSorted((a, b) => a - b)[Math.floor(values.length / 2)];
+  const glyphs = values.filter(v => Math.abs(v - background) > 48);
+  if (glyphs.length === 0) return 'dark-on-light';
+  return glyphs.reduce((a, b) => a + b, 0) / glyphs.length > background ? 'light-on-dark' : 'dark-on-light';
+}
 /** Linear stretch so the darkest 2% become black and the lightest 2% white. */
 function stretch(raster) {
   const values = new Uint8Array(raster.width * raster.height);
@@ -837,7 +859,7 @@ async function readPage(page, engine, options) {
     const second = await recheckRun(engine, scanImage, run, {
       ...rules,
       ...recheck
-    });
+    }, printPolarity(page.original.raster, originalDpi, run));
     if (second.reading === null) continue;
     claims.found[r] = second.reading;
     rechecks.cleared++;
@@ -943,5 +965,5 @@ 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, claimWords, compareTexts, cosine, createTesseractEngine, diacriticsMap, dice, foldConfusables, foldDiacritics, jaccard, jaroWinkler, judgeRun, judgeRuns, levenshtein, levenshteinSimilarity, matchWords, normaliseText, ocrPages, recheckRun, tokenise, wordDistance, wordRecall };
+export { DEFAULT_NORMALISE, DEFAULT_RECHECK_PASSES, DEFAULT_TESSERACT_OPTIONS, claimWords, compareTexts, cosine, createTesseractEngine, diacriticsMap, dice, foldConfusables, foldDiacritics, jaccard, jaroWinkler, judgeRun, judgeRuns, levenshtein, levenshteinSimilarity, matchWords, normaliseText, ocrPages, printPolarity, recheckRun, tokenise, 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 { claimWords, DEFAULT_RECHECK_PASSES, judgeRun, judgeRuns, matchWords, recheckRun } from './page-reading/index.js';
-export type { Claims, MatchOptions, Recheck, RecheckOptions, RecheckPass, Reference, Verdict, WordMatch } from './page-reading/index.js';
+export { claimWords, DEFAULT_RECHECK_PASSES, judgeRun, judgeRuns, matchWords, printPolarity, recheckRun } from './page-reading/index.js';
+export type { Claims, MatchOptions, PrintPolarity, 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';
 export { diacriticsMap, foldConfusables, foldDiacritics } from './text-normalisation/index.js';
 //# sourceMappingURL=index.d.ts.map

package/dist/src/page-reading/index.d.ts

@@ -3,6 +3,6 @@ export { claimWords, judgeRun, judgeRuns, matchWords } from './match-words.use-c
 export type { Claims, MatchOptions, Reference, Verdict, WordMatch } from './match-words.use-case.js';
 export { ocrPages } from './ocr-pages.use-case.js';
 export type { OcrOptions, OcrReport, PageOcr, PlacedText, PositionedText, ReadablePage, RunReading, SideText, TextDifference } from './ocr-report.contract.js';
-export { DEFAULT_RECHECK_PASSES, recheckRun } from './recheck-run.use-case.js';
-export type { Recheck, RecheckOptions, RecheckPass } from './recheck-run.use-case.js';
+export { DEFAULT_RECHECK_PASSES, printPolarity, recheckRun } from './recheck-run.use-case.js';
+export type { PrintPolarity, Recheck, RecheckOptions, RecheckPass } from './recheck-run.use-case.js';
 //# sourceMappingURL=index.d.ts.map

package/dist/src/page-reading/recheck-run.use-case.d.ts

@@ -9,7 +9,10 @@ import type { MatchOptions, Reference } from './match-words.use-case.js';
  * a shaded bar - is at the mercy of the engine's layout analysis and of one
  * resolution. Cropped to itself, enlarged further, read as a single line (or
  * word, or as digits only when it is a figure), light-on-dark text inverted,
- * contrast stretched, it often reads cleanly. Measured on real scans at 120
+ * contrast stretched, it often reads cleanly. Whether the text is light on a
+ * dark bar is taken from the original, where it is certain, rather than
+ * guessed from the scan: a white total on a purple bar can scan as pale lilac
+ * on paler lilac, too light overall to look like a dark bar. Measured on real scans at 120
  * and 144 dpi, cropped multi-pass reading confirmed 96-97% of every printed
  * value on the page; at 93 dpi, a third.
  *
@@ -34,6 +37,8 @@ export interface RecheckOptions {
     agree?: number;
 }
 export declare const DEFAULT_RECHECK_PASSES: readonly RecheckPass[];
+/** Which way the original prints a run: dark text on light, or light text on a dark bar. */
+export type PrintPolarity = 'dark-on-light' | 'light-on-dark';
 export interface Recheck {
     cleared: boolean;
     /** An agreeing reading when cleared; `null` otherwise. */
@@ -45,5 +50,11 @@ export interface Recheck {
 export declare function recheckRun(engine: OcrEngine, image: {
     raster: Raster;
     dpi: number;
-}, run: Reference, options: MatchOptions & RecheckOptions): Promise<Recheck>;
+}, run: Reference, options: MatchOptions & RecheckOptions, polarity?: PrintPolarity): Promise<Recheck>;
+/**
+ * Which way the original prints a run, read off its own crisp rendering: the
+ * glyphs are the pixels far from the background, and the background is most of
+ * the box.
+ */
+export declare function printPolarity(raster: Raster, dpi: number, run: Reference): PrintPolarity;
 //# sourceMappingURL=recheck-run.use-case.d.ts.map

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@scanmate/ocr",
-  "version": "0.0.2",
+  "version": "0.0.3",
   "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",
@@ -51,6 +51,6 @@
   },
   "devDependencies": {
     "@scanmate/align": "^0.0.2",
-    "@scanmate/extract": "^0.0.2"
+    "@scanmate/extract": "^0.0.3"
   }
 }