GoodScanner

• TODO

  • OpenCV Android

    • https://webnautes.tistory.com/923

  • Gaussian Blurring Image noise

  • OpenCV Document Edge Detection

    • https://www.pyimagesearch.com/2014/09/01/build-kick-ass-mobile-document-scanner-just-5-minutes/

  • Edge Detection Document Contour(Outline)

  • OpenCV Perspective Transform

    • https://www.pyimagesearch.com/2014/08/25/4-point-opencv-getperspective-transform-example/
  • • De-noise image Text Recognition

  • Model document docx(?)

Text Recognition Details

• Text Recognition

  • StillImageActivity.kt
    • onGlobalLayout(), onActivityResult() tryReloadAndDetectInImage()
    • tryReloadAndDetectInImage()
      • imageUri image BitmapUtils.getBitmapFromContentUri() image Bitmap
      • Bitmap size Text Recognition AI model input size resize resizedBitmap .
      • imageProcessor!!.processBitmap(resizedBitmap, graphicOverlay) resizedBitmap graphicOverlay .
        • imageProcessor = TextRecognitionProcessor(this, KoreanTextRecognizerOptions.Builder().build())
        • processBitmap() VisionProcessorBase.kt
  • VisionProcessorBase.kt
    • processBitmap()
      • requestDetectInImage()
    • requestDetectInImage()
      • tryReloadAndDetectInImage() resizedBitmap input(image: InputImage)
      • setUpListener
        • detectInImage(image)
        • detectInImage() OnSuccessListener OnFailureListener (addOnSuccessListener, addOnFailureListener)
          • [email protected](results, graphicOverlay)
          • [email protected](e)
  • TextRecognitionProcessor.kt
    • detectInImage()
      • VisionProcessorBase.kt abstract fun TextRecognitionProcessor.kt
      • textRecognizer.process(image)
      • textRecognizer: TextRecognizer = TextRecognition.getClient(textRecognizerOptions)
    • process()
      • AI model
      • JNI Native code OnSuccessListner onSuccess() OnFailureListner onFailure()
    • onSuccess()
      • graphicOverlay TextGraphic(graphicOverlay, text, shouldGroupRecognizedTextInBlocks) text
      • text add TextGraphic.kt draw()

• Text Recognition

  • VisionProcessorBase.kt
    • [email protected](results, graphicOverlay)
  • TextRecognitionProcessor.kt
    • onSuccess(text: Text, graphicOverlay: com.good.scanner.GraphicOverlay)
  • TextGraphic.kt
    • private val text: Text
    • Text class .

• Text class

  • text.text:

  • text.textBlocks detection textblock for

    for (textBlock in text.textBlocks) { // Renders the text at the bottom of the box.
      Log.d(TAG, "TextBlock text is: " + textBlock.text)
      Log.d(TAG, "TextBlock boundingbox is: " + textBlock.boundingBox)
      Log.d(TAG, "TextBlock cornerpoint is: " + Arrays.toString(textBlock.cornerPoints))
    

  • textblock lines for

  for (line in textBlock.lines) {
    Log.d(TAG, "Line text is: " + line.text)
    Log.d(TAG, "Line boundingbox is: " + line.boundingBox)
    Log.d(TAG, "Line cornerpoint is: " + Arrays.toString(line.cornerPoints))
  

  • line elements for

  for (element in line.elements) {
    Log.d(TAG, "Element text is: " + element.text)
    Log.d(TAG, "Element boundingbox is: " + element.boundingBox)
    Log.d(TAG, "Element cornerpoint is: " + Arrays.toString(element.cornerPoints))
    Log.d(TAG, "Element language is: " + element.recognizedLanguage)
  

1. StillImageActivity.kt imageProcessor.!!processBitmap() Text Recognition bitmap OpenCV bitmap(image) bitmap

   • bitmap pdf !

   • bitmap processBitmap parameter text recognition !

2. Text Recognition visualization

   • web textblock visualization

   • Best docx layout visualization(?)

     • Text size bounding box layout
     • bounding box image layout