tf.raw_ops.EncodeJxl

JPEG XL-encode an image.

tf.raw_ops.EncodeJxl(
    image, quality=95, effort=7, name=None
)

image is a rank-N uint8 or uint16 Tensor of shape batch_dims + [height, width, channels] where channels is:

  • 1: for grayscale.
  • 3: for RGB.
  • 4: for RGBA.

quality is mapped internally to a Butteraugli distance, using the same mapping as cjxl -q: 100.0 maps to distance 0.0 (lossless modular encoding), 95.0 maps to distance 0.5 (default lossy VarDCT encoding), 90.0 maps to distance 1.0, and 0.0 maps to distance 25.0. Note that quality factors are not comparable across codecs; a given quality will not produce the same file size or visual quality as the same value passed to tf.io.encode_jpeg.

image A Tensor. Must be one of the following types: uint8, uint16. Images to encode. Rank N >= 3 with shape batch_dims + [height, width, channels].
quality An optional float. Defaults to 95. JPEG-style quality factor in [0.0, 100.0]. Defaults to 95.0 to match encode_jpeg. 100.0 encodes losslessly, 90.0 is visually lossless, and lower values compress more.
effort An optional int. Defaults to 7. Effort level from 1 (fastest) to 9 (slowest/best compression).
name A name for the operation (optional).

A Tensor of type string.