how to write convert yuv array to rgb array?

Question

I have original data which is yuv420p frame data bytes. I want to convert it to rgb data using numpy and scipy. Here is my code:

    yuv = np.frombuffer(data, dtype='uint8')
    y = yuv[:1920*1080].reshape(1080, 1920)
    v = yuv[1920*1080::2].reshape(540, 960)
    u = yuv[1920*1080+1::2].reshape(540, 960)

    u = ndimage.zoom(u, 2, order=0)
    v = ndimage.zoom(v, 2, order=0)

    y = y.reshape((y.shape[0], y.shape[1], 1))
    u = u.reshape((u.shape[0], u.shape[1], 1))
    v = v.reshape((v.shape[0], v.shape[1], 1))

    yuv = np.concatenate((y, u, v), axis=2)

    yuv[:, :, 0] = yuv[:, :, 0].clip(16, 235).astype(yuv.dtype) - 16
    yuv[:, :, 1:] = yuv[:, :, 1:].clip(16, 240).astype(yuv.dtype) - 128

    A = np.array([[1.164, 0.000, 1.793],
                  [1.164, -0.213, -0.533],
                  [1.164, 2.112, 0.000]])

    rgb = np.dot(yuv, A.T).clip(0, 255).astype('uint8')

I used PIL to open this output rgb array and the image wasn't the way I expected it to be.

Is there anying wrong with my code?? Or is it wrong with my data?

Answer 1

The problem is in the format of the original video. yuv420 means that the chroma components (u and v) have one forth of the spacial resolution od the y component. That is the reason why you are seeing four 'little' images overlayed on top of your y component. Therefore, you have to upsample the u and v components by a factor of two, both in the horizontal and vertical directions, so that each component can match the spatial resolution of the y component.