Skip to main content

Python package to parse and provide access to headers and data streams in Amira(R) files

Project description


ahds (Amira (R) header and data streams) is a Python package to parse and handle Amira (R) files. It was developed to facilitate reading of Amira (R) files as part of the EMDB-SFF toolkit.

Use Cases

  • Detect and parse Amira (R) headers and return structured data
  • Decode data (HxRLEByte, HxZip)
  • Easy extensibility to handle previously unencountered data streams

ahds was written and is maintained by Paul K. Korir.


  • SimpleParse (tested with 2.1.1 - install from source from the SimpleParse site; FAILs with 2.2.0)
  • numpy (tested with 1.11.2)
  • scikit-image (tested with 0.11.3)


Copyright 2017 EMBL - European Bioinformatics Institute

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

Unless required by applicable law or agreed to in writing,
software distributed under the License is distributed on an
either express or implied. See the License for the specific
language governing permissions and limitations under the License.

Future Plans

  • Write out valid Amira (R) files

Background and Definitions

ahds presently handles two types of Amira (R) files:

  • AmiraMesh files, which typically but not necessarily have a .am extension, and
  • HyperSurface files, which have .surf and represent an older filetype.

Both file types consist of two parts:

  • a header, and
  • one or more data streams.

Headers are structured in a modified VRML-like syntax and differ between AmiraMesh and HyperSurface files in some of the keywords used.

A data stream is a sequence of encoded bytes either referred to in the header by some delimiter (usually @<data_stream_index>, where <data_stream_index> is an integer) or a set of structural keywords (e.g. Vertices, Patches) expected in a predefined sequence.

Headers in Detail

AmiraMesh and HyperSurface headers can be divided into four main sections:

  • designation
  • definitions
  • parameters, and
  • data pointers.

The designation is the first line and conveys several important details about the format and structure of the file such as:

  • filetype (either AmiraMesh or HyperSurface)
  • dimensionality (3D)
  • version (a decimal number e.g. 2.1
  • extra format data e.g. <hxsurface> specifying that an AmiraMesh file will contain HyperSurface data

A series of definitions follow that refer to data found in the data pointer sections that either begin with the word ‘define’ or have ‘n’ prepended to a variable. For example:

define Lattice 862 971 200


nVertices 85120

This is followed by grouped parameters enclosed in a series of braces beginning with the word ‘Parameters’. Various parameters are then enclosed each beginning with the name of that group of parameters e.g. ‘Materials’

Parameters {
        # grouped parameters
        Material {
                # the names of various materials with attributes
                Exterior {
                        id 0
                Inside {
                        id 1,
                        Color 0 1 1,
                        Transparency 0.5
        Patches {
        # patch attributes
                InnerRegion “Inside”,
                OuterRegion “Exterior”,
                BoundaryID 0,
                BranchingPoints 0
        # inline parameters
        GridSize <value>,

The most important set of parameters are materials as these specify colours and identities of distinct segments/datasets within the file.

Finally, AmiraMesh files list a set of data pointers that point to data labels within the file together with additional information to decode the data. We refer to these as data streams because they consist of continuous streams of raw byte data that need to be decoded. Here is an example of data pointers that refer to the location of 3D surface primitives:

Vertices { float[3] Vertices } @1
TriangleData { int[7] Triangles } @2
Patches-0 { int Patches-0 } @3

These refer to three raw data streams each found beginning with the delimiter @<number>. Data stream one (@1) is called Vertices and consists of float triples, two is called TriangleData and has integer 7-tuples and three called Patches- is a single integer (the number of patches). In some cases the data pointer contains the data encoding for the corresponding data pointer.

Lattice { byte Labels } @1(HxByteRLE,234575740)

which is a run-length encoded data stream of the specified length, while

Lattice { byte Data } @1(HxZip,919215)

contains zipped data of the specified length.

Data Streams in Detail

AmiraMesh data streams are very simple. They always have a start delimiter made of @ with an index that identifies the data stream. A newline character separates the delimiter with the data stream proper which is either plain ASCII or a binary stream (raw, zipped or encoded).

HyperSurface data streams structured to have the following sections:

# Header
Vertices <nvertices>
# vertices data stream

NBranchingPoints <nbranching_points>
NVerticesOnCurves <nvertices_on_curves>
BoundaryCurves <nboundary_curves>
Patches <npatches>
InnerRegion <inner_region_name>
OuterRegion <outer_region_name>
BoundaryID <boundary_id>
BranchingPoints <nbranching_points>
Triangles <ntriangles>
# triangles data stream
} # repeats for as <npatches> times

HyperSurface data streams can be either plain ASCII or binary.

ahds Modules

ahds has three main modules:

  • ahds.grammar specifies an EBNF grammar
  • ahds.header
  • ahds.data_stream

These modules are tied into a user-level class called AmiraFile that does all the work for you.

>>> from ahds import AmiraFile
>>> # read an AmiraMesh file
>>> af = AmiraFile('am/')
>>> af.header
<AmiraHeader with 4 bytes>
>>> # empty data streams
>>> af.data_streams
>>> print af.data_streams
>>> # we have to explicitly read to get the data streams
>>> af.data_streams
<class 'ahds.data_stream.DataStreams'> object with 13 stream(s): 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13
>>> for ds in af.data_streams:
...   print ds
<class 'ahds.data_stream.AmiraMeshDataStream'> object of 2,608 bytes
<class 'ahds.data_stream.AmiraMeshDataStream'> object of 2,608 bytes
<class 'ahds.data_stream.AmiraMeshDataStream'> object of 2,608 bytes
<class 'ahds.data_stream.AmiraMeshDataStream'> object of 2,608 bytes
<class 'ahds.data_stream.AmiraMeshDataStream'> object of 2,608 bytes
<class 'ahds.data_stream.AmiraMeshDataStream'> object of 2,608 bytes
<class 'ahds.data_stream.AmiraMeshDataStream'> object of 2,608 bytes
<class 'ahds.data_stream.AmiraMeshDataStream'> object of 2,608 bytes
<class 'ahds.data_stream.AmiraMeshDataStream'> object of 2,608 bytes
<class 'ahds.data_stream.AmiraMeshDataStream'> object of 2,608 bytes
<class 'ahds.data_stream.AmiraMeshDataStream'> object of 2,608 bytes
<class 'ahds.data_stream.AmiraMeshDataStream'> object of 2,608 bytes
<class 'ahds.data_stream.AmiraMeshDataStream'> object of 2,608 bytes
# we get the n-th data stream using the index/key notation
>>> af.data_streams[1].encoded_data
'1 \n2 \n3 \n'
>>> af.data_streams[1].decoded_data
[1, 2, 3]
>>> af.data_streams[2].encoded_data
'69 \n120 \n116 \n101 \n114 \n105 \n111 \n114 \n0 \n73 \n110 \n115 \n105 \n100 \n101 \n0 \n109 \n111 \n108 \n101 \n99 \n117 \n108 \n101 \n0 \n'
>>> af.data_streams[2].decoded_data
[69, 120, 116, 101, 114, 105, 111, 114, 0, 73, 110, 115, 105, 100, 101, 0, 109, 111, 108, 101, 99, 117, 108, 101, 0]
>>> # read an HyperSurface file
>>> af = AmiraFile('surf/')
>>> af.data_streams
<class 'ahds.data_stream.DataStreams'> object with 5 stream(s): Patches, NBranchingPoints, BoundaryCurves, Vertices, NVerticesOnCurves
# HyperSurface files have pre-set data streams
>>> af.data_streams['Vertices'].decoded_data[:10]
[(560.0, 243.0, 60.96875), (560.0, 242.9166717529297, 61.0), (559.5, 243.0, 61.0), (561.0, 243.0, 60.95833206176758), (561.0, 242.5, 61.0), (561.0384521484375, 243.0, 61.0), (559.0, 244.0, 60.94444274902344), (559.0, 243.5, 61.0), (558.9722290039062, 244.0, 61.0), (560.0, 244.0, 60.459999084472656)]


This module describes the header grammar for Amira (R) (AmiraMesh and HyperSurface) files and so depends on simpleparse Python package. It defines a single class (AmiraDispatchProcessor) and four functions.

AmiraDispatchProcessor is a subclass of simpleparse.dispatchprocessor which implements the core functionality required to use the grammar. Each grammar token has a corresponding method defined on this class which determines how the data associated with that token will be rendered. Data can be rendered as a single or multimap, string, number, or in custom format.

  • ahds.grammar.get_parsed_data(fn, *args, **kwargs) is the user-level function that takes a filename and returns structured parsed data. It depends on the other three functions defined:
  • ahds.grammar.detect_format(fn, format_bytes=50, verbose=False) returns either AmiraMesh or HyperSurface given a file name and arguments,
  • get_header(fn, file_format, header_bytes=20000, verbose=False) returns the header portion based on the file format determined by detect_format(…), and
  • parse_header(data, verbose=False) converts the raw header data returned by get_header(…) into a structured header based on AmiraDispatchProcessor.


This module converts the structured header from the ahds.grammar module into an object with the sections of the header (designation, definitions, parameters and data pointers) and corresponding structured data available as attributes. That is it converts the header


define Lattice 862 971 200

Parameters {
    Materials {
        Exterior {
            Id 1
        Inside {
            Color 0.64 0 0.8,
            Id 2
        Mitochondria {
            Id 3,
            Color 0 1 0
        Mitochondria_ {
            Id 4,
            Color 1 1 0
        mitochondria__ {
            Id 5,
            Color 0 0.125 1
        NE {
            Id 6,
            Color 1 0 0
    Content "862x971x200 byte, uniform coordinates",
    BoundingBox 0 13410.7 0 15108.4 1121.45 4221.01,
    CoordType "uniform"

Lattice { byte Labels } @1(HxByteRLE,4014522)

into an AmiraHeader object.

>>> from ahds.header import AmiraHeader
>>> amira_header = AmiraHeader.from_file('am/')
>>> amira_header.designation.attrs
['filetype', 'dimension', 'format', 'version', 'extra_format']
>>> amira_header.designation.filetype
>>> amira_header.designation.dimension
>>> amira_header.designation.format
>>> amira_header.definitions.attrs
>>> amira_header.definitions.Lattice
[862, 971, 200]
>>> amira_header.parameters.attrs
['Materials', 'Content', 'BoundingBox', 'CoordType']
>>> amira_header.parameters.Materials.attrs
['Exterior', 'Inside', 'Mitochondria', 'Mitochondria_', 'mitochondria__', 'NE']
>>> amira_header.parameters.Materials.Exterior.attrs
>>> amira_header.parameters.Materials.Exterior.Id
>>> amira_header.parameters.Content
'"862x971x200 byte, uniform coordinates",'
>>> amira_header.parameters.BoundingBox
[0, 13410.7, 0, 15108.4, 1121.45, 4221.01]
>>> amira_header.parameters.CoordType
>>> amira_header.data_pointers.attrs
>>> amira_header.data_pointers.data_pointer_1.attrs
['pointer_name', 'data_format', 'data_dimension', 'data_type', 'data_name', 'data_index', 'data_length']
>>> amira_header.data_pointers.data_pointer_1.pointer_name
>>> amira_header.data_pointers.data_pointer_1.data_format
>>> amira_header.data_pointers.data_pointer_1.data_dimension
>>> amira_header.data_pointers.data_pointer_1.data_type
>>> amira_header.data_pointers.data_pointer_1.data_name
>>> amira_header.data_pointers.data_pointer_1.data_index
>>> amira_header.data_pointers.data_pointer_1.data_length

This module consists of two main classes: ahds.header.AmiraHeader is the user-level class and ahds.header.Block which is a container class for a block of structured data from an Amira (R) header.

AmiraHeader has one constructor: AmiraHeader.from_file(fn, *args, **kwargs) which takes an Amira (R) file by name and arguments and returns an AmiraHeader object with all attributes set as described above. Alternatively, one can use the initiator form to pass structured data directly: AmiraHeader(parsed_data) which returns an AmiraHeader object configured appropriately.

  • The raw data structured data is available as read-only property: AmiraHeader.raw_header
  • Internally the AmiraHeader class implements a set of private methods which individually load the four data sections (designation, definitions, parameters, and data pointers).

The Block class is a container class which converts structured groups to attributes and has two main attributes:

  • provides the name of the current block
  • Block.attrs provides the attributes available on this Block
>>> amira_header.designation.attrs
['filetype', 'dimension', 'format', 'version', 'extra_format']
>>> amira_header.designation.format
A given Materials block has two special features:
Block.ids returns the list of ids for all materials. This is important when decoding HxByteRLE compressed data
Block[id] returns the material for the given id using index notation.
>>> amira_header.parameters.Materials.ids
[1, 2, 3, 4, 5, 6]
>>> amira_header.parameters.attrs
['Materials', 'Content', 'BoundingBox', 'CoordType']
# ids attribute is only available for ‘Material’ blocks within ‘parameters’ section
>>> amira_header.parameters.Content.ids
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: 'str' object has no attribute 'ids'
# we can get the name of a material of the given id
>>> amira_header.parameters.Materials[4].name


This is most complex module implementing a hierarchy of classes describing various data streams within Amira (R) files. It has 22 classes and five functions


There are three categories of classes:

  • A user-level class that encapsulates (2) below.
  • Classes describing Amira (R) data streams
    • Classes describing AmiraMesh data streams
    • Classes describing HyperSurface data streams
  • Data conversion classes (AmiraMesh only)
    • Classes abstracting images
    • Classes abstracting contours

The user-level DataStreams class is the preferred way to use the module. It takes the name of an Amira (R) file and encapsulates an iterator of data streams.

>>> from ahds import data_stream
>>> data_streams = data_stream.DataStreams('am/')
>>> data_streams
<class 'ahds.data_stream.DataStreams'> object with 2 stream(s): 1, 2
>>> for ds in data_streams:
...   print ds
<class 'ahds.data_stream.AmiraMeshDataStream'> object of 968,909 bytes
<class 'ahds.data_stream.AmiraMeshDataStream'> object of 968,909 bytes


The functions implemented in this module decode data streams.

  • ahds.data_stream.hxbyterle_decode decodes HxByteRLE data streams
  • ahds.data_stream.hxzip_decode(data_size, data) unzips zlib-compressed data streams
  • ahds.data_stream.unpack_binary(data_pointer, definitions, data) unpacks the structured data stream according to the attributes specified in the data’s data pointer
  • ahds.data_stream.unpack_ascii(data) converts rows of ASCII data into numerical data

Classes in Detail

DataStreams class

The following attributes are available on objects of this class:

  • ahds.data_stream.DataStreams.file - filename of Amira (R) file
  • ahds.data_stream.DataStreams.header - an object of class ahds.header.AmiraHeader encapsulating the header data in four sections (designation, definitions, parameters, and data pointers)
  • ahds.data_stream.DataStreams.filetype - the filetype as specified in (ii) above.
  • ahds.data_stream.DataStreams.stream_data - all raw data from the file (including the header)
  • len(DataStreams) - the number of data streams contained
  • ahds.data_stream.DataStreams[<index>] - returns the data stream of the index specified (as defined in the data_pointers section of the header object
Classes describing Amira (R) data streams

The following diagrams illustrates the hierarchy of classes:

Classes describing Amira (R) data streams

  • ahds.data_stream.AmiraDataStream is the base class for all data stream classes and defines the following attributes:
    • ahds.data_stream.AmiraDataStream.header - an ahds.header.AmiraHeader object
    • ahds.data_stream.AmiraDataStream.data_pointer - the ahds.header.AmiraHeader.data_pointers.data_pointer_X for this data stream
    • ahds.data_stream.AmiraDataStream.stream_data - the raw file data
    • ahds.data_stream.AmiraDataStream.encoded_data - the encoded data for this stream; None for VoidDataStream subclasses
    • ahds.data_stream.AmiraDataStream.decoded_data - the decoded data for this stream; None for VoidDataStream subclasses
    • ahds.data_stream.AmiraDataStream.decoded_length - the number of items (tuples, integers) in decoded data

The two main subclasses of AmiraDataStream are ahds.data_stream.AmiraMeshDataStream, which is a concrete class representing all AmiraMesh data streams, and ahds.data_stream.AmiraHxSurfaceDataStream, which abstractly defines HyperSurface data streams.

There are two main AmiraHxSurfaceDataStream subclasses:

  • ahds.data_stream.VoidDataStream represents AmiraHxSurfaceDataStream data streams that only have a name and value but no actual encoded data (on the following line). There are two subclasses:
    • ahds.data_stream.NamedDataStream subclasses have a strings after data stream name. The two concrete subclasses are:
      • ahds.data_stream.PatchesInnerRegionDataStream for the name of an inner region of a patch (see PatchesDataStream), and
      • ahds.data_stream.PatchesOuterRegionDataStream for corresponding name of the outer region of a patch.
    • ahds.data_stream.ValuedDataStream have an integer value after the data stream name. The three concrete subclasses are:
      • ahds.data_stream.PatchesBoundaryIDDataStream hold the boundary ID of a patch,
      • ahds.data_stream.PatchesBranchingPointsDataStream stores the number of branching points, and
      • ahds.data_stream.PatchesDataStream with the number of patches, which is a special ValueDataStream that contains an iterable of patches each containing a Patches<X>DataStream objects.
    • ahds.data_stream.LoadedDataStream represent AmiraHxSurfaceDataStream data streams that have a name, a value and encoded data. The two main concrete subclasses are:
      • ahds.data_stream.VerticesDataStream represents data streams with float-triples, and
      • ahds.data_stream.PatchesTrianglesDataStream represents data streams within a patch with triples of 1-based indices (triangles) of vertices specified in the VerticesDataStream.
Conversion classes

There are two groups of conversion classes which only apply to (some) AmiraMesh data streams: Conversion classes

  • Image conversion classes consist of a image container class ImageSet and an Image class. ImageSet objects that can be iterated to give Image objects are returned from the AmiraMeshDataStream.to_images() method call.
>>> # decode the data stream to images
>>> images = ds[1].to_images()
>>> images
<ImageSet with 200 images>
>>> for image in images:
...     print image
<Image with dimensions (971, 862)>
<Image with dimensions (971, 862)>
<Image with dimensions (971, 862)>
<Image with dimensions (971, 862)>
<Image with dimensions (971, 862)>
  • Contour conversion classes convert individual images into sets of contours (ContourSet) iterable as individual Contours objects. They are obtained from calls to the Image.as_contours property. Furthermore, the Image.as_segments property call returns a dictionary of the corresponding ContourSet object indexed by the z plane.
>>> # contours per image
>>> # the dictionary key is the Amira Id for the segment (the Id of the Material)
>>> # a segment can have several non-overlapping contours (or polylines)
>>> for image in images:
...     print image.as_contours
{2: <class 'ahds.data_stream.ContourSet'> with 15 contours, 3: <class 'ahds.data_stream.ContourSet'> with 3 contours, 5: <class 'ahds.data_stream.ContourSet'> with 2 contours}
{2: <class 'ahds.data_stream.ContourSet'> with 18 contours, 3: <class 'ahds.data_stream.ContourSet'> with 3 contours, 5: <class 'ahds.data_stream.ContourSet'> with 2 contours}
{2: <class 'ahds.data_stream.ContourSet'> with 15 contours, 3: <class 'ahds.data_stream.ContourSet'> with 1 contours, 5: <class 'ahds.data_stream.ContourSet'> with 3 contours}
{2: <class 'ahds.data_stream.ContourSet'> with 15 contours, 3: <class 'ahds.data_stream.ContourSet'> with 1 contours, 5: <class 'ahds.data_stream.ContourSet'> with 3 contours}

>>> # separate individual segments
>>> images.segments
{1: {110: <class 'ahds.data_stream.ContourSet'> with 1 contours}, 2: {0: <class 'ahds.data_stream.ContourSet'> with 15 contours, 1: <class 'ahds.data_stream.ContourSet'> with 18 contours, ..., 198: <class 'ahds.data_stream.ContourSet'> with 3 contours, 199: <class 'ahds.data_stream.ContourSet'> with 3 contours}}

Project details

Release history Release notifications

This version
History Node


History Node


History Node


History Node


History Node


History Node


History Node


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Filename, size & hash SHA256 hash help File type Python version Upload date
ahds-0.1.9.tar.gz (29.1 kB) Copy SHA256 hash SHA256 Source None Jan 23, 2018

Supported by

Elastic Elastic Search Pingdom Pingdom Monitoring Google Google BigQuery Sentry Sentry Error logging CloudAMQP CloudAMQP RabbitMQ AWS AWS Cloud computing DataDog DataDog Monitoring Fastly Fastly CDN DigiCert DigiCert EV certificate StatusPage StatusPage Status page