OutOfFocusQRReader

OutOfFocusQRReader is a library for reading QR Code from out-of-focus images on Android.

This library uses the OpenCV library.

“QR Code” is a registered trademark of DENSO WAVE INCORPORATED.

Install

Download OpenCV

Download the latest version of the OpenCV Android library from the official OpenCV web page.

https://opencv.org/releases/

Extract zip and move SDK folder

Extract the files the downloaded zip file and move the SDK folder to your project root folder.

Install OpenCV SDK

Edit the settings.gradle file in your project and sync your project:

// add
include ':sdk'

Install OutOfFocusQRReader

Edit the build.gradle file in your project:

repositories {
    mavenCentral()
}

dependencies {
    // OpenCV sdk
    implementation project(':sdk')

    // core
    implementation 'com.github.soranakk.oofqrreader:oofqrreader:1.0.0'
}

Usage

Create ImageData

Example using Camera2API

dependencies {
    ...

    implementation 'com.github.soranakk.oofqrreader:image-converter-android-camera:1.0.0'
}

val imageConverter = Camera2ApiImageConverter()
ImageReader.newInstance(width, height, ImageFormat.YUV_420_888, 1)
    .apply {
        setOnImageAvailableListener({ reader ->
            reader?.acquireLatestImage()?.let { image ->
                val imageData = imageConverter.convertImage(image)
                ... read QR Code
                image.close()
            }
        }, workerHandler)

Example using CameraXAPI

dependencies {
    ...

    implementation 'com.github.soranakk.oofqrreader:image-converter-androidx-camera:1.0.0'
}

val imageConverter = CameraXApiImageConverter()
val imageAnalysis = ImageAnalysis.Builder()
  .setTargetRotation(Surface.ROTATION_0)
  .setTargetResolution(Size(1280, 720))
  .setBackpressureStrategy(ImageAnalysis.STRATEGY_KEEP_ONLY_LATEST)
  .build()
  .also {
      it.setAnalyzer(cameraExecutor, ImageAnalysis.Analyzer { image ->
          val imageData = imageConverter.convertImage(image)
          ... read QR Code
          image.close()
      })
  }

Create QRCodeDecoder

Example using MLKit:

dependencies {
    ...

    implementation 'com.github.soranakk.oofqrreader:decoder-mlkit:1.0.0'
    implementation 'com.google.mlkit:barcode-scanning:16.1.1'
}

val mlKitDecoder = MLKitDecoder()

Example using ZXing:

dependencies {
    ...

    implementation 'com.github.soranakk.oofqrreader:decoder-zxing:1.0.0'
    implementation "com.google.zxing:core:3.4.1"
}

val zxingDecoder = ZxingDecoder()

If your app require Android minSdkVersion less than 24, see https://github.com/zxing/zxing/issues/1170 and demoApp's build.gradle

Example using OpenCV:

dependencies {
    ...

    implementation 'com.github.soranakk.oofqrreader:decoder-opencv:1.0.0'
}

val openCVDecoder = OpenCVDecoder()

Read QR code

val qrReader = MultiFilterQRCodeReader(someDecoder)
val result = qrReader.detectAndRead(image)

See demoApp for more info.

Appendix

How to improve accuracy

Estimate the area where a QR code exist

Estimate the area where a QR code exists by using the large contrast between the QR code and the background.

1. Make the input image grayscale
2. Binarize with Otsu's method
3. Use Imgproc.erode to thicken the dots in a QR code and turn them into black squares.
4. Use Imgproc.findContours to get the black square area

Adjust the image

Further improve the accuracy by processing the image of the estimated area using various Image Filters. The features of each ImageFilter implemented are as follows.

GaussianThresholdFilter

Binarize with a weighted average using Gaussian as a threshold at a fixed block size.

Input image

Estimate the area where a QR code exists

After applying the image filter

ThresholdOtsuFilter

Binarize Otsu's method.

Input image

Estimate the area where a QR code exists

After applying the image filter

OverexposureFilter

A filter for reading overexposed QR codes by setting a threshold so that even almost white areas become black. Most input images output a black image, but under certain circumstances this filter works well. For example, displaying a QR code on a smartphone will result in overexposure due to the strong backlight of the smartphone.

Input image

Estimate the area where a QR code exists

After applying the image filter