react-native-rectangle-doc-scanner 3.244.0 → 3.246.0

package/android/src/main/kotlin/com/reactnativerectangledocscanner/CameraController.kt

@@ -48,6 +48,7 @@ class CameraController(
     private var previewSize: Size? = null
     private var analysisSize: Size? = null
     private var captureSize: Size? = null
+    private var sensorOrientation: Int = 0
 
     private var yuvReader: ImageReader? = null
     private var jpegReader: ImageReader? = null
@@ -207,6 +208,7 @@ class CameraController(
             val characteristics = cameraManager.getCameraCharacteristics(cameraId)
             val streamConfigMap = characteristics.get(CameraCharacteristics.SCALER_STREAM_CONFIGURATION_MAP)
                 ?: return
+            sensorOrientation = characteristics.get(CameraCharacteristics.SENSOR_ORIENTATION) ?: 0
 
             val viewAspect = if (previewView.height == 0) {
                 1.0
@@ -370,11 +372,35 @@ class CameraController(
         val rotationDegrees = computeRotationDegrees()
         val imageWidth = image.width
         val imageHeight = image.height
+        val nv21 = try {
+            imageToNv21(image)
+        } catch (e: Exception) {
+            Log.e(TAG, "[CAMERA2] Failed to read image buffer", e)
+            try {
+                image.close()
+            } catch (closeError: Exception) {
+                Log.w(TAG, "[CAMERA2] Failed to close image", closeError)
+            }
+            analysisInFlight.set(false)
+            return
+        } finally {
+            try {
+                image.close()
+            } catch (e: Exception) {
+                Log.w(TAG, "[CAMERA2] Failed to close image", e)
+            }
+        }
+
         val inputImage = try {
-            InputImage.fromMediaImage(image, rotationDegrees)
+            InputImage.fromByteArray(
+                nv21,
+                imageWidth,
+                imageHeight,
+                rotationDegrees,
+                InputImage.IMAGE_FORMAT_NV21
+            )
         } catch (e: Exception) {
             Log.e(TAG, "[CAMERA2] Failed to create InputImage", e)
-            image.close()
             analysisInFlight.set(false)
             return
         }
@@ -388,7 +414,9 @@ class CameraController(
                 val mlBox = best?.boundingBox
                 val rectangle = when {
                     mlBox == null -> null
-                    shouldRefineWithOpenCv() -> refineWithOpenCv(image, rotationDegrees, mlBox) ?: boxToRectangle(mlBox)
+                    shouldRefineWithOpenCv() ->
+                        refineWithOpenCv(nv21, imageWidth, imageHeight, rotationDegrees, mlBox)
+                            ?: boxToRectangle(mlBox)
                     else -> boxToRectangle(mlBox)
                 }
 
@@ -400,11 +428,6 @@ class CameraController(
                 Log.e(TAG, "[CAMERA2] ML Kit detection failed", e)
             }
             .addOnCompleteListener {
-                try {
-                    image.close()
-                } catch (e: Exception) {
-                    Log.w(TAG, "[CAMERA2] Failed to close image", e)
-                }
                 analysisInFlight.set(false)
             }
     }
@@ -454,9 +477,6 @@ class CameraController(
     }
 
     private fun computeRotationDegrees(): Int {
-        val cameraId = selectCameraId() ?: return 0
-        val characteristics = cameraManager.getCameraCharacteristics(cameraId)
-        val sensorOrientation = characteristics.get(CameraCharacteristics.SENSOR_ORIENTATION) ?: 0
         val displayRotation = displayRotationDegrees()
         return if (useFrontCamera) {
             (sensorOrientation + displayRotation) % 360
@@ -482,7 +502,7 @@ class CameraController(
         val preview = previewSize ?: return
         if (viewWidth == 0f || viewHeight == 0f) return
 
-        val rotation = displayRotationDegrees()
+        val rotation = computeRotationDegrees()
         val bufferWidth = if (rotation == 90 || rotation == 270) preview.height.toFloat() else preview.width.toFloat()
         val bufferHeight = if (rotation == 90 || rotation == 270) preview.width.toFloat() else preview.height.toFloat()
 
@@ -557,16 +577,21 @@ class CameraController(
         return true
     }
 
-    private fun refineWithOpenCv(image: Image, rotationDegrees: Int, mlBox: Rect): Rectangle? {
+    private fun refineWithOpenCv(
+        nv21: ByteArray,
+        imageWidth: Int,
+        imageHeight: Int,
+        rotationDegrees: Int,
+        mlBox: Rect
+    ): Rectangle? {
         return try {
-            val nv21 = imageToNv21(image)
-            val uprightWidth = if (rotationDegrees == 90 || rotationDegrees == 270) image.height else image.width
-            val uprightHeight = if (rotationDegrees == 90 || rotationDegrees == 270) image.width else image.height
+            val uprightWidth = if (rotationDegrees == 90 || rotationDegrees == 270) imageHeight else imageWidth
+            val uprightHeight = if (rotationDegrees == 90 || rotationDegrees == 270) imageWidth else imageHeight
             val expanded = expandRect(mlBox, uprightWidth, uprightHeight, 0.2f)
             val openCvRect = DocumentDetector.detectRectangleInYUVWithRoi(
                 nv21,
-                image.width,
-                image.height,
+                imageWidth,
+                imageHeight,
                 rotationDegrees,
                 expanded
             )

package/android/src/main/kotlin/com/reactnativerectangledocscanner/DocumentScannerView.kt

@@ -50,6 +50,9 @@ class DocumentScannerView(context: ThemedReactContext) : FrameLayout(context), L
     private var lastDetectionTimestamp: Long = 0L
     private var isCapturing = false
     private var isDetaching = false
+    private var lastDetectedRectangle: Rectangle? = null
+    private var lastDetectedImageWidth = 0
+    private var lastDetectedImageHeight = 0
 
     // Coroutine scope for async operations
     private val scope = CoroutineScope(Dispatchers.Main + SupervisorJob())
@@ -186,6 +189,12 @@ class DocumentScannerView(context: ThemedReactContext) : FrameLayout(context), L
         }
         lastDetectionTimestamp = now
 
+        if (rectangle != null && imageWidth > 0 && imageHeight > 0) {
+            lastDetectedRectangle = rectangle
+            lastDetectedImageWidth = imageWidth
+            lastDetectedImageHeight = imageHeight
+        }
+
         val rectangleOnScreen = if (rectangle != null && width > 0 && height > 0) {
             DocumentDetector.transformRectangleToViewCoordinates(rectangle, imageWidth, imageHeight, width, height)
         } else {
@@ -313,12 +322,21 @@ class DocumentScannerView(context: ThemedReactContext) : FrameLayout(context), L
             // Detect rectangle in captured image
             val bitmap = BitmapFactory.decodeFile(imageFile.absolutePath)
                 ?: throw IllegalStateException("decode_failed")
-            val detectedRectangle = try {
+            var detectedRectangle = try {
                 DocumentDetector.detectRectangle(bitmap)
             } catch (e: Exception) {
                 Log.w(TAG, "Rectangle detection failed, using original image", e)
                 null
             }
+            if (detectedRectangle == null && lastDetectedRectangle != null && lastDetectedImageWidth > 0 && lastDetectedImageHeight > 0) {
+                detectedRectangle = scaleRectangleToBitmap(
+                    lastDetectedRectangle!!,
+                    lastDetectedImageWidth,
+                    lastDetectedImageHeight,
+                    bitmap.width,
+                    bitmap.height
+                )
+            }
 
             // Process image with detected rectangle
             val shouldCrop = detectedRectangle != null && stableCounter > 0
@@ -389,9 +407,7 @@ class DocumentScannerView(context: ThemedReactContext) : FrameLayout(context), L
     }
 
     private fun sendPictureTakenEvent(data: WritableMap) {
-        val event = Arguments.createMap().apply {
-            merge(data)
-        }
+        val event = data.toHashMap().toWritableMap()
         themedContext.getJSModule(RCTEventEmitter::class.java)
             .receiveEvent(id, "onPictureTaken", event)
     }
@@ -582,6 +598,27 @@ class DocumentScannerView(context: ThemedReactContext) : FrameLayout(context), L
             }
         }
     }
+
+    private fun scaleRectangleToBitmap(
+        rectangle: Rectangle,
+        srcWidth: Int,
+        srcHeight: Int,
+        dstWidth: Int,
+        dstHeight: Int
+    ): Rectangle {
+        if (srcWidth == 0 || srcHeight == 0) return rectangle
+        val scaleX = dstWidth.toDouble() / srcWidth.toDouble()
+        val scaleY = dstHeight.toDouble() / srcHeight.toDouble()
+        fun mapPoint(point: Point): Point {
+            return Point(point.x * scaleX, point.y * scaleY)
+        }
+        return Rectangle(
+            mapPoint(rectangle.topLeft),
+            mapPoint(rectangle.topRight),
+            mapPoint(rectangle.bottomLeft),
+            mapPoint(rectangle.bottomRight)
+        )
+    }
 }
 
 /**

package/dist/DocScanner.js

@@ -334,7 +334,9 @@ exports.DocScanner = (0, react_1.forwardRef)(({ onCapture, overlayColor = DEFAUL
             captureOriginRef.current = 'auto';
         },
     }), [capture]);
-    const overlayPolygon = detectedRectangle?.rectangleOnScreen ?? detectedRectangle?.rectangleCoordinates ?? null;
+    const overlayPolygon = react_native_1.Platform.OS === 'android'
+        ? detectedRectangle?.rectangleOnScreen ?? null
+        : detectedRectangle?.rectangleOnScreen ?? detectedRectangle?.rectangleCoordinates ?? null;
     const overlayIsActive = autoCapture ? isAutoCapturing : (detectedRectangle?.stableCounter ?? 0) > 0;
     const detectionThreshold = autoCapture ? minStableFrames : 99999;
     return (react_1.default.createElement(react_native_1.View, { style: styles.container },

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "react-native-rectangle-doc-scanner",
-  "version": "3.244.0",
+  "version": "3.246.0",
   "description": "Native-backed document scanner for React Native with customizable overlays.",
   "license": "MIT",
   "main": "dist/index.js",

package/src/DocScanner.tsx

@@ -452,7 +452,10 @@ export const DocScanner = forwardRef<DocScannerHandle, Props>(
       [capture],
     );
 
-    const overlayPolygon = detectedRectangle?.rectangleOnScreen ?? detectedRectangle?.rectangleCoordinates ?? null;
+    const overlayPolygon =
+      Platform.OS === 'android'
+        ? detectedRectangle?.rectangleOnScreen ?? null
+        : detectedRectangle?.rectangleOnScreen ?? detectedRectangle?.rectangleCoordinates ?? null;
     const overlayIsActive = autoCapture ? isAutoCapturing : (detectedRectangle?.stableCounter ?? 0) > 0;
 
   const detectionThreshold = autoCapture ? minStableFrames : 99999;