✨ add support for crop & split operations (#457)

Author: sebastianMindee

src/image/extractedImage.ts

@@ -15,11 +15,15 @@ import { loadOptionalDependency } from "@/dependency/index.js";
  */
 export class ExtractedImage {
   public buffer: Buffer;
-  protected internalFileName: string;
+  public filename: string;
+  public pageId?: number;
+  public elementId?: number;
 
-  protected constructor(buffer: Uint8Array, fileName: string) {
+  constructor(buffer: Uint8Array, fileName: string, pageId?: number, elementId?: number) {
     this.buffer = Buffer.from(buffer);
-    this.internalFileName = fileName;
+    this.filename = fileName;
+    this.pageId = pageId;
+    this.elementId = elementId;
   }
 
   /**
@@ -104,7 +108,7 @@ export class ExtractedImage {
   asSource(): BufferInput {
     return new BufferInput({
       buffer: this.buffer,
-      filename: this.internalFileName,
+      filename: this.filename,
     });
   }
 }

src/image/imageExtractor.ts

@@ -1,59 +1,107 @@
+import { loadOptionalDependency } from "@/dependency/index.js";
+import { MindeeImageError } from "@/errors/index.js";
+import { getMinMaxX, getMinMaxY, Polygon } from "@/geometry/index.js";
+import { adjustForRotation } from "@/geometry/polygonUtils.js";
+import { ExtractedImage } from "@/image/extractedImage.js";
+import { LocalInputSource } from "@/input/index.js";
+import { logger } from "@/logger.js";
+import { createPdfFromInputSource } from "@/pdf/pdfOperation.js";
+import { rasterizePage } from "@/pdf/pdfUtils.js";
 // eslint-disable-next-line @typescript-eslint/ban-ts-comment
 // @ts-ignore
 import type * as pdfLibTypes from "@cantoo/pdf-lib";
-import { getMinMaxX, getMinMaxY, Polygon } from "@/geometry/index.js";
-import { adjustForRotation } from "@/geometry/polygonUtils.js";
-import { loadOptionalDependency } from "@/dependency/index.js";
 
 let pdfLib: typeof pdfLibTypes | null = null;
 
 async function getPdfLib(): Promise<typeof pdfLibTypes> {
   if (!pdfLib) {
     const pdfLibImport = await loadOptionalDependency<typeof pdfLibTypes>(
-      "@cantoo/pdf-lib", "Text Embedding"
+      "@cantoo/pdf-lib", "Image Extraction"
     );
     pdfLib = (pdfLibImport as any).default || pdfLibImport;
   }
   return pdfLib!;
 }
 
+
+/**
+ * Extracts elements from a PDF document based on a list of bounding boxes.
+ * @param inputSource The input source to extract from.
+ * @param polygonsPerPage List of polygons to extract from per page.
+ * @param quality JPEG quality of extracted images.
+ */
+export async function extractImagesFromPolygon(
+  inputSource: LocalInputSource,
+  polygonsPerPage: Map<number, Polygon[]>,
+  quality?: number
+) {
+  const allExtractedImages: ExtractedImage[] = [];
+  const pdfDoc = await createPdfFromInputSource(inputSource);
+
+  for (const [pageId, polygons] of polygonsPerPage) {
+    logger.debug(`Extracting images from page ${pageId}`);
+    const pdfPage = pdfDoc.getPage(pageId);
+    const extractions = (await extractFromPage(pdfPage, polygons, true, quality));
+    const extractedImages = extractions.map(
+      (v, i) => new ExtractedImage(v, inputSource.filename + `_page${pageId}-${i}.jpg`, pageId, i)
+    );
+    allExtractedImages.push(...extractedImages);
+  }
+  return allExtractedImages;
+}
+
 /**
  * Extracts elements from a page based off of a list of bounding boxes.
  *
  * @param pdfPage PDF Page to extract from.
  * @param polygons List of coordinates to pull the elements from.
+ * @param asImage Whether to return the extracted elements as images.
+ * @param quality JPEG quality of extracted images, given as number between 0 and 1.
  */
 export async function extractFromPage(
   pdfPage: pdfLibTypes.PDFPage,
-  polygons: Polygon[]
+  polygons: Polygon[],
+  asImage: boolean = false,
+  quality?: number,
 ) {
   const pdfLib = await getPdfLib();
   const { width, height } = pdfPage.getSize();
-  const extractedElements :Uint8Array[] = [];
-  // Manual upscale.
-  // Fixes issues with the OCR.
-  const qualityScale = 300/72;
+  const extractedElements: Uint8Array[] = [];
+  if (quality && (quality < 0)) {
+    throw new MindeeImageError("Quality must be a number between 0 and 1");
+  }
+  if (quality && quality > 1) {
+    logger.warn("Quality is greater than 1, this operation will apply a manual upscale on the output." +
+      " Use only if you know what you are doing.");
+  }
+  const qualityScale = quality ?? 1;
   const orientation = pdfPage.getRotation().angle;
 
+  const sourceDoc = pdfPage.doc;
+  const pageIndex = sourceDoc.getPages().indexOf(pdfPage);
+
   for (const origPolygon of polygons) {
-    const polygon = adjustForRotation(origPolygon, orientation);
+    logger.debug(`Extracting image with polygon: ${origPolygon.toString()}`);
 
     const tempPdf = await pdfLib.PDFDocument.create();
 
+    const [copiedPage] = await tempPdf.copyPages(sourceDoc, [pageIndex]);
+
+    const polygon = adjustForRotation(origPolygon, orientation);
+
     const newWidth = width * (getMinMaxX(polygon).max - getMinMaxX(polygon).min);
     const newHeight = height * (getMinMaxY(polygon).max - getMinMaxY(polygon).min);
-    const cropped = await tempPdf.embedPage(pdfPage, {
+
+    const cropped = await tempPdf.embedPage(copiedPage, {
       left: getMinMaxX(polygon).min * width,
       right: getMinMaxX(polygon).max * width,
       top: height - (getMinMaxY(polygon).min * height),
       bottom: height - (getMinMaxY(polygon).max * height),
     });
 
-    // Determine the final page dimensions based on orientation
     let finalWidth: number;
     let finalHeight: number;
     if (orientation === 90 || orientation === 270) {
-      // For 90/270 rotations, swap width and height
       finalWidth = newHeight * qualityScale;
       finalHeight = newWidth * qualityScale;
     } else {
@@ -62,15 +110,14 @@ export async function extractFromPage(
     }
 
     const samplePage = tempPdf.addPage([finalWidth, finalHeight]);
-
     samplePage.drawRectangle({
       x: 0,
       y: 0,
       width: finalWidth,
       height: finalHeight,
+      color: pdfLib.rgb(1, 1, 1),
     });
 
-    // Draw the cropped page with rotation applied
     if (orientation === 0) {
       samplePage.drawPage(cropped, {
         width: newWidth * qualityScale,
@@ -102,7 +149,13 @@ export async function extractFromPage(
       });
     }
 
-    extractedElements.push(await tempPdf.save());
+    const pdfBuffer = Buffer.from(await tempPdf.save());
+    if (asImage) {
+      extractedElements.push(await rasterizePage(pdfBuffer, 0, 100));
+    } else {
+      extractedElements.push(pdfBuffer);
+    }
   }
+
   return extractedElements;
 }

src/pdf/extractedPdf.ts

@@ -0,0 +1,71 @@
+import path from "node:path";
+import { BufferInput, MIMETYPES } from "@/input/index.js";
+import { MindeeError } from "@/errors/index.js";
+import { Buffer } from "node:buffer";
+import { writeFile } from "fs/promises";
+import { logger } from "@/logger.js";
+import { writeFileSync } from "node:fs";
+
+export class ExtractedPdf {
+  public readonly buffer: Buffer;
+  public readonly filename: string;
+  public readonly pageCount: number;
+
+  constructor(pdfData: Buffer<ArrayBufferLike>, filename: string, pageCount: number) {
+    this.buffer = pdfData;
+    this.filename = filename;
+    this.pageCount = pageCount;
+  }
+
+  /**
+   * Saves the document to a file.
+   *
+   * @param outputPath Path to save the file to.
+   */
+  async saveToFileAsync(outputPath: string) {
+    const fileExt = path.extname(outputPath).toLowerCase();
+    if (fileExt !== ".pdf" && !MIMETYPES.has(fileExt)) {
+      outputPath += ".pdf";
+    }
+
+    try {
+      await writeFile(path.resolve(outputPath), this.buffer);
+      logger.info(`File saved successfully to ${path.resolve(outputPath)}.`);
+    } catch (e) {
+      if (e instanceof TypeError) {
+        throw new MindeeError("Invalid path/filename provided.");
+      } else {
+        throw e;
+      }
+    }
+  }
+
+  /**
+   * Saves the document to a file synchronously.
+   * @param outputPath
+   */
+  saveToFile(outputPath: string){
+    try {
+      writeFileSync(path.resolve(outputPath), this.buffer);
+      logger.info(`File saved successfully to ${path.resolve(outputPath)}.`);
+    } catch (e) {
+      if (e instanceof TypeError) {
+        throw new MindeeError("Invalid path/filename provided.");
+      } else {
+        throw e;
+      }
+    }
+  }
+
+  /**
+   * Return the file as a Mindee-compatible BufferInput source.
+   *
+   * @returns A BufferInput source.
+   */
+  asSource(): BufferInput {
+    return new BufferInput({
+      buffer: this.buffer,
+      filename: this.filename,
+    });
+  }
+}

src/pdf/pdfCompressor.ts

@@ -1,15 +1,10 @@
 import { logger } from "@/logger.js";
-import tmp from "tmp";
-import * as fs from "node:fs";
-// eslint-disable-next-line @typescript-eslint/ban-ts-comment
-// @ts-ignore
-import type * as popplerTypes from "node-poppler";
 // eslint-disable-next-line @typescript-eslint/ban-ts-comment
 // @ts-ignore
 import type * as pdfLibTypes from "@cantoo/pdf-lib";
 import { compressImage } from "@/image/index.js";
 import { loadOptionalDependency } from "@/dependency/index.js";
-import { ExtractedPdfInfo, extractTextFromPdf, hasSourceText } from "./pdfUtils.js";
+import { ExtractedPdfInfo, extractTextFromPdf, hasSourceText, rasterizePage } from "./pdfUtils.js";
 
 let pdfLib: typeof pdfLibTypes | null = null;
 
@@ -159,7 +154,7 @@ async function compressPagesWithQuality(
     const page = pdfDoc.getPages()[i];
     const rasterizedPage = await rasterizePage(pdfData, i + 1, imageQuality);
     const compressedImage = await compressImage(
-      Buffer.from(rasterizedPage, "binary"), imageQuality
+      rasterizedPage, imageQuality
     );
     if (!disableSourceText) {
       await addTextToPdfPage(page, extractedText);
@@ -260,48 +255,6 @@ async function getFontFromName(fontName: string): Promise<pdfLibTypes.PDFFont> {
   return font;
 }
 
-/**
- * Rasterizes a PDF page.
- *
- * @param pdfData Buffer representation of the entire PDF file.
- * @param index Index of the page to rasterize.
- * @param quality Quality to apply during rasterization.
- */
-async function rasterizePage(
-  pdfData: Buffer, index: number, quality = 85
-): Promise<string> {
-  const popplerImport = await loadOptionalDependency<typeof popplerTypes>(
-    "node-poppler", "Image Processing"
-  );
-  const poppler = (popplerImport as any).default || popplerImport;
-  const popplerInstance = new poppler.Poppler();
-  const tmpPdf = tmp.fileSync();
-  const tempPdfPath = tmpPdf.name;
-  const antialiasOption: "fast" | "best" | "default" | "good" | "gray" | "none" | "subpixel" = "best";
-  try {
-    await fs.promises.writeFile(tempPdfPath, pdfData);
-    const options = {
-      antialias: antialiasOption,
-      firstPageToConvert: index,
-      lastPageToConvert: index,
-      jpegFile: true,
-      jpegOptions: `quality=${quality}`,
-      singleFile: true
-    };
-
-    const jpegBuffer = await popplerInstance.pdfToCairo(tempPdfPath, undefined, options);
-
-    await fs.promises.unlink(tempPdfPath);
-
-    return jpegBuffer;
-  } catch (error) {
-    logger.error("Error rasterizing PDF:", error);
-    throw error;
-  } finally {
-    tmpPdf.removeCallback();
-  }
-}
-
 /**
  * Performs linear interpolation between two numbers.
  * @param start The starting value.