pyimcom.compress.i24

Functions to compress float arrays to 24 bit integers.

Functions

lsbf_fwd: Least significant bits moved first.
lsbf_rev: Inverse of lsbf_fwd.
diff_fwd: Replaces an image with an array of differences.
diff_rev: Inverse of diff_fwd.
smallnum_fwd: Re-maps numbers of small absolute value to be near 0 when unsigned.
smallnum_rev: Inverse of smallnum_fwd.
i24compress: Compresses an array using the ‘I24’ scheme.
i24decompress: Decompresses an array using the ‘I24’ scheme.

Classes

I24Cube: Object that can be reformatted as floating point, integer, or compressed integer in the ‘I24’ scheme.

Attributes

i24_recognized_schemes

Classes

I24Cube

Class to compress a float cube to int24 with scaling, and reconstruct the original image.

Functions

`lsbf_fwd`(im)	Takes a 2D uint8 image and swaps the bits so that the least significant
`lsbf_rev`(im)	Inverse of the bit-swapping function lsbf_fwd.
`diff_fwd`(im, bitkeep)	Replace an int32 image with differences of the same shape.
`diff_rev`(im, bitkeep)	Reconstruct an image from differences. Inverse function of diff_fwd.
`smallnum_fwd`(im, bitkeep)	Remapping to package small differences of either sign near 0.
`smallnum_rev`(im, bitkeep)	Reconstructs images that have been re-mapped. Inverse of smallnum_fwd.
`i24compress`(im, scheme, pars)	Compresses an image.
`i24decompress`(im, scheme, pars[, overflow])	Decompresses an image.

Module Contents

i24_recognized_schemes = ['I24A', 'I24B'][source]

lsbf_fwd(im)[source]

Takes a 2D uint8 image and swaps the bits so that the least significant bit goes first, then the next, etc. and the most significant bit goes last.

If given a 3D image, then does each 2D slice. The intent would be to use this on objects with a shape of (3,ny,nx); it will work for any number of slices, but it will be slow if the number of slices (3 in this case) is large.

Parameters:: im (np.array of uint8) – 2D or 3D image.
Returns:: out_im – 2D or 3D bit-swapped image. Same shape as im.
Return type:: np.array of uint8

See also

lsbf_rev: Inverse function.

lsbf_rev(im)[source]

Inverse of the bit-swapping function lsbf_fwd.

Parameters:: im (np.array of uint8) – 2D or 3D image.
Returns:: out_im – 2D or 3D bit-swapped image. Same shape as im.
Return type:: np.array of uint8

See also

lsbf_fwd: Forward function.

diff_fwd(im, bitkeep)[source]

Replace an int32 image with differences of the same shape.

Parameters:

im (np.array of int32) – A 2D input image.
bitkeep (int) – Number of bits to keep. The maximum is 31.

Returns:

The image of differences, same shape as im.

Return type:

np.array of int32

See also

diff_rev: Inverse function.

diff_rev(im, bitkeep)[source]

Reconstruct an image from differences. Inverse function of diff_fwd.

Parameters:

im (np.array of int32) – A 2D input image of differences.
bitkeep (int) – Number of bits to keep. The maximum is 31.

Returns:

The original image, same shape as im.

Return type:

np.array of int32

See also

diff_fwd: Forward function.

smallnum_fwd(im, bitkeep)[source]

Remapping to package small differences of either sign near 0.

Works on an int32 image with bitkeep bits used and negative numbers rolling over as -j --> 2**bitkeep-j.

Parameters:

im (np.array of int32) – A 2D input image.
bitkeep (int) – Number of bits to keep. The maximum is 31.

Returns:

Re-mapped image, same shape as im.

Return type:

np.array of int32

See also

smallnum_rev: Inverse function.

smallnum_rev(im, bitkeep)[source]

Reconstructs images that have been re-mapped. Inverse of smallnum_fwd.

Parameters:

im (np.array of int32) – A 2D input image that has been re-mapped.
bitkeep (int) – Number of bits to keep. The maximum is 31.

Returns:

Original image, same shape as im.

Return type:

np.array of int32

See also

smallnum_fwd: Forward function.

class I24Cube(inarray, pars, overflow=None)[source]

Class to compress a float cube to int24 with scaling, and reconstruct the original image.

Parameters:

inarray (np.array) – The input array. Options are * 2D float32 (original image) * 2D int32 (intermediate step: leading byte not used) * 3D uint8 (compressed form)
pars (dict) – Parameters for the compression scheme. The possible keys are VMIN, VMAX, SOFTBIAS, DIFF, ALPHA, BITKEEP, and REORDER.
overflow (astropy.io.fits.BinTableHDU or None, optional) – Overflow table (y,x,value). Needed if a compressed image is given as input.

ny

Height of image

Type:: int

nx

Width of image

Type:: int

pars[source]

Compression parameters specified when constructed.

Type:: dict

data[source]

Current image data.

Type:: np.array

mode

Current image type. One of ‘float32’, ‘int32’, or ‘uint8’.

Type:: str

vmin[source]

Minimum of compression range.

Type:: float

vmax[source]

Maximum of compression range.

Type:: float

alpha

Power law index of compression (1=linear).

Type:: float

bitkeep

Number of bits to keep (maximum=24).

Type:: int

reorder

Implement bit reordering?

Type:: bool

softbias

Integer in [0,2**24) to add (to avoid slight negative fluctuations being 111111). If -1, uses smallnum compression instead.

Type:: int

diff

Use difference of successive pixels?

Type:: bool

overflow[source]

Overflow table (y,x,value).

Type:: astropy.io.fits.BinTableHDU or None

reorder = use bit reordering? (boolean)

__init__()[source]: Constructor.

to_mode()[source]: Change image format to a different (often compressed) storage mode.

pars[source]

data[source]

vmin[source]

vmax[source]

overflow = None[source]

to_mode(mode)[source]

Converts to the given mode.

Parameters:: mode (str) – Options are ‘float32’, ‘int32’, and ‘uint8’.
Return type:: None

i24compress(im, scheme, pars)[source]

Compresses an image.

Parameters:

im (np.array of uint8) – 2D or 3D image.
scheme (str) – Compression scheme (right now supports ‘I24A’, ‘I24B’).
pars (dict) – Parameters to pass to compression algorithm.

Returns:

data (np.array) – The compressed data cube.
overflow (astropy.io.fits.BinTableHDU) – The table of values that overflowed the quantization range.

See also

pyimcom.compress.i24.I24Cube: Compression class; see for possible keys in pars.

i24decompress(im, scheme, pars, overflow=None)[source]

Decompresses an image.

Parameters:

im (np.array of uint8) – 2D or 3D image.
scheme (str) – Compression scheme (right now supports ‘I24A’, ‘I24B’).
pars (dict) – Parameters to pass to compression algorithm.
overflow (astropy.io.fits.BinTableHDU or None, optional) – Overflow table (y,x,value). Needed if a compressed image is given as input.

Returns:

data – The de-compressed data cube.

Return type:

np.array

See also

pyimcom.compress.i24.I24Cube: Compression class; see for possible keys in pars.