tf.image.yuv_to_rgb

View source on GitHub

Converts one or more images from YUV to RGB.

View aliases

Compat aliases for migration

See Migration guide for more details.

tf.compat.v1.image.yuv_to_rgb

tf.image.yuv_to_rgb(
    images
)

Outputs a tensor of the same shape as the images tensor, containing the RGB value of the pixels. The output is only well defined if the Y value in images are in [0,1], U and V value are in [-0.5,0.5].

As per the above description, you need to scale your YUV images if their pixel values are not in the required range. Below given example illustrates preprocessing of each channel of images before feeding them to yuv_to_rgb.

yuv_images = tf.random.uniform(shape=[100, 64, 64, 3], maxval=255)
last_dimension_axis = len(yuv_images.shape) - 1
yuv_tensor_images = tf.truediv(
    tf.subtract(
        yuv_images,
        tf.reduce_min(yuv_images)
    ),
    tf.subtract(
        tf.reduce_max(yuv_images),
        tf.reduce_min(yuv_images)
     )
)
y, u, v = tf.split(yuv_tensor_images, 3, axis=last_dimension_axis)
target_uv_min, target_uv_max = -0.5, 0.5
u = u * (target_uv_max - target_uv_min) + target_uv_min
v = v * (target_uv_max - target_uv_min) + target_uv_min
preprocessed_yuv_images = tf.concat([y, u, v], axis=last_dimension_axis)
rgb_tensor_images = tf.image.yuv_to_rgb(preprocessed_yuv_images)

Args
images	2-D or higher rank. Image data to convert. Last dimension must be size 3.

Returns
images	tensor with the same shape as images.