OpenCV on Android: practices and tips


Transfer MAT objects from Android to NDK

The main idea is to use the address of an MAT object in order to manipulate the data.

Basically, we have a function playing as a bridge between Java APIs and NDK:

public native void function_name(long matAddress);

To call the function, we use Mat’s address by calling getNativeObjAddr(). All computations in NDK will affect the content of MAT in both Java and NDK layers.

In the NDK code, to use cv::Mat object regarding java Mat, we can use static_cast:

Mat& im = *(static_cast<Mat*>(addrImg));


There is a huge different in image channels between Java OpenCV and NDK OpenCV. When we decode the path or file to a bitmap in Java, we have to convert the bitmap to ARGB_8888 color channel, otherwise it does not work. Actually, it also can work on RGB_565 but for some reasons I can not remember, I always use ARGB_8888 in the project.

Bitmap bm32_image = bm.copy(Bitmap.Config.ARGB_8888, true);

To manipulate the image correctly in NDK, we should covert it to the normal RGB channel, otherwise sometimes we get some bugs making us frustrating.

// convert ARGB_8888 to RGB
Mat im = new Mat();
Mat rgb_im = new Mat();
Utils.bitmapToMat(bm32_image, im);
Imgproc.cvtcolor(im, rgb_im, Imgproc.COLOR_RGBA2RGB, 3);


Other practice is to put some assertions in NDK code to make sure that we the use the correct format for the input.

OpenCV’s Camera

OpenCV supports 3 types of camera:


  1. Create the layout of the camera. For example, we can put the following lines to the Activity xml file:
  1. Next, in the activity which controls the camera, we have to implement required methods from the CvCameraViewLisener2. There 3 methods are:

Prior to processing the camera, we need to initialize the variable holding callbacks of three aforementioned methods:

private CameraBridgeViewBase mOpenCVCamera;

On the onCreate method:

    mOpenCVCamera = (CameraBridgeViewBase) findViewById(;

Next, we implement 3 required methods:

    public void onCameraViewStarted(int width, int height) {
        // initialize images, variables, settings

    public void onCameraViewStopped() {
        // release the resources

    public Mat onCameraFrame(CameraBridgeViewBase.CvCameraViewFrame inputFrame) {
        // retrieve the frame from `inputFrame`
        // - the grayscale frame by imputFrame.gray()
        // - the RGBA frame by inputFrame.rgba()
        Mat im = inputFrame.rgba();

        // do things
        // postprocess: convert back to the RGBA image
        return im; // `im` will show in the UI

Noting that OpenCV’s Camera is not able to set the portrait mode. One workaround is to turn the Activity to landscape by putting the following line inside tag Activity in AndroidManifest.xml


Data Manipulation

unsigned char Mat

It is troublesome when we want to assign a value of 255 to an unsigned char Mat because this language does not support unsigned char as a primitive type. One workaround is to allocate a 16S Mat, manipulate on that matrix, and finally convert to 8U.

Point2f and Point

To convert MatOfPoint to MatOfPoint2f, we use the constructor:

MatOfPoint matofpoint = new MatOfPoint(matofpoint2f.toArray());

Accessing the pixel values

In order to retrieve and assign pixel value, we use the getter/setter from Mat.

short[] pixel = new short[nchannels];

m.get(i, j, pixel); // retrieve pixel values at (i, j)
// in this example, the pixel has 3 channels
pixel[0] = 255;
pixel[1] = 0;
pixel[2] = 125;
m.set(i, j, pixel);