Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

Decoding a YUV image in Python OpenCV

Tags:

python

opencv

I have a YUV420_SP_NV21 image represented as a byte array (no headers), taken from an Android preview frame, and I need to decode it into a RGB image. I've done this in Android apps before, using Java and OpenCV4Android:

convert_mYuv = new Mat(height + height / 2, width, CvType.CV_8UC1);
convert_mYuv.put( 0, 0, data );
Imgproc.cvtColor( convert_mYuv, convert_mRgba, type, channels );

I've tried doing the same in Python:

nmp = np.array(byteArray, dtype=np.ubyte) RGBMatrix = cv2.cvtColor(nmp, cv2.COLOR_YUV420P2RGB)

, but RGBMatrix remains None.

I'm aware of the possibility to do this myself, I'm familiar with the formula, but would be realy happy to do this the OpenCV way. How can this be done?

I've also tried cv2.imdecode(), but it failed too, possibly because of me miss-using it.

like image 415
Ysch Avatar asked May 08 '26 13:05

Ysch


1 Answers

Its been a while since I solved this issue, but I hadn't the time to update this question.

I ended up extracting the YUV420_SP_NV21 byte array to a full scaled YUV image (1280x720 in my case) using numpy, then converting it to RGB using OpenCV.

    import cv2 as cv2        # OpenCV import
    def YUVtoRGB(byteArray):

        e = 1280*720
        Y = byteArray[0:e]
        Y = np.reshape(Y, (720,1280))

        s = e
        V = byteArray[s::2]
        V = np.repeat(V, 2, 0)
        V = np.reshape(V, (360,1280))
        V = np.repeat(V, 2, 0)

        U = byteArray[s+1::2]
        U = np.repeat(U, 2, 0)
        U = np.reshape(U, (360,1280))
        U = np.repeat(U, 2, 0)

        RGBMatrix = (np.dstack([Y,U,V])).astype(np.uint8)
        RGBMatrix = cv2.cvtColor(RGBMatrix, cv2.COLOR_YUV2RGB, 3)

My raw data was YUV420_SP_NV21, which is decode like this: YY...YVUVUVU...VU, but other YUV formats will require subtle changes to the code.

like image 158
Ysch Avatar answered May 10 '26 03:05

Ysch



Donate For Us

If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!