pydicom.pixels.utils.pixel_array

pydicom.pixels.utils.pixel_array(src: str | PathLike[str] | BinaryIO | Dataset, *, ds_out: Dataset | None = None, specific_tags: list[int] | None = None, index: int | None = None, raw: bool = False, decoding_plugin: str = '', **kwargs: Any) → np.ndarray[source]

Return decoded pixel data from src as ndarray.

Added in version 3.0.

Warning

This function requires NumPy and may require the installation of additional packages to perform the actual pixel data decompression. See the pixel data decompression documentation for more information.

Memory Usage

To minimize memory usage src should be the path to the dataset or a file-like object containing the dataset.

Processing

The following processing operations on the raw pixel data are always performed:

  • Natively encoded bit-packed pixel data for a bits allocated of 1 will be unpacked.

  • Natively encoded pixel data with a photometric interpretation of "YBR_FULL_422" will have it’s sub-sampling removed.

  • The output array will be reshaped to the specified dimensions.

  • JPEG-LS or JPEG 2000 encoded data whose signedness doesn’t match the expected pixel representation will be converted to match.

If raw = False (the default) then the following processing operation will also be performed:

Examples

Read a DICOM dataset and return the entire pixel data:

from pydicom import dcmread
from pydicom.pixels import pixel_array

ds = dcmread("path/to/dataset.dcm")
arr = pixel_array(ds)

Return the entire pixel data from a dataset while minimizing memory usage:

from pydicom.pixels import pixel_array

arr = pixel_array("path/to/dataset.dcm")

Return the 3rd frame of a dataset containing at least 3 frames while minimizing memory usage:

from pydicom.pixels import pixel_array

with open("path/to/dataset.dcm", "rb") as f:
    arr = pixel_array(f, index=2)  # 'index' starts at 0
Parameters:

  • src (str | PathLike[str] | file-like | pydicom.dataset.Dataset) –

  • ds_out (pydicom.dataset.Dataset, optional) – A Dataset that will be updated with the non-retired group 0x0028 image pixel module elements and the group 0x0002 file meta information elements from the dataset in src. Only available when `src` is a path or file-like.

  • specific_tags (list[int | pydicom.tag.BaseTag], optional) – A list of additional tags from the dataset in src to be added to the ds_out dataset.

  • index (int | None, optional) – If None (default) then return an array containing all the frames in the pixel data, otherwise return only the frame from the specified index, which starts at 0 for the first frame.

  • raw (bool, optional) – If True then return the decoded pixel data after only minimal processing (see the processing section above). If False (default) then additional processing may be applied to convert the pixel data to it’s most commonly used form (such as converting from YCbCr to RGB).

  • decoding_plugin (str, optional) – The name of the decoding plugin to use when decoding compressed pixel data. If no decoding_plugin is specified (default) then all available plugins will be tried and the result from the first successful one returned. For information on the available plugins for each decoder see the API documentation.

  • **kwargs – Optional keyword parameters for controlling decoding, please see the decoding options documentation for more information.

Returns:

One or more frames of decoded pixel data with shape:

  • (rows, columns) for single frame, single sample data

  • (rows, columns, samples) for single frame, multi-sample data

  • (frames, rows, columns) for multi-frame, single sample data

  • (frames, rows, columns, samples) for multi-frame, multi-sample data

A writeable ndarray is returned by default. For native transfer syntaxes with view_only=True a read-only ndarray will be returned.

Return type:

numpy.ndarray