tf.contrib.image.dense_image_warp

tf.contrib.image.dense_image_warp(
    image,
    flow,
    name='dense_image_warp'
)

Defined in tensorflow/contrib/image/python/ops/dense_image_warp.py.

Image warping using per-pixel flow vectors.

Apply a non-linear warp to the image, where the warp is specified by a dense flow field of offset vectors that define the correspondences of pixel values in the output image back to locations in the source image. Specifically, the pixel value at output[b, j, i, c] is images[b, j - flow[b, j, i, 0], i - flow[b, j, i, 1], c].

The locations specified by this formula do not necessarily map to an int index. Therefore, the pixel value is obtained by bilinear interpolation of the 4 nearest pixels around (b, j - flow[b, j, i, 0], i - flow[b, j, i, 1]). For locations outside of the image, we use the nearest pixel values at the image boundary.

Args:

  • image: 4-D float Tensor with shape [batch, height, width, channels].
  • flow: A 4-D float Tensor with shape [batch, height, width, 2].
  • name: A name for the operation (optional).

    Note that image and flow can be of type tf.half, tf.float32, or tf.float64, and do not necessarily have to be the same type.

Returns:

A 4-D float Tensor with shape[batch, height, width, channels] and same type as input image.

Raises:

  • ValueError: if height < 2 or width < 2 or the inputs have the wrong number of dimensions.