modules.image_assembly

KPF Image Assembly module.

Assembles a raw L0 readout into a single L1 full-frame image (overscan correction, amplifier stitching). It applies no calibrations: bias/dark/flat are handled downstream by ImageProcessing following the standard CCD reduction sequence (bias subtraction, then exposure-scaled dark subtraction, then flat division), gated per file type by the masters modules.

class kpfpipe.modules.image_assembly.ImageAssembly(l0_obj, config=None)

Bases: object

This class performs CCD-level processing to convert L0 data to L1.

Operations include:

  • orienting amplifier channels

  • applying gain conversion (ADU –> photo-electrons)

  • measuring read noise

  • subtracting overscan bias

  • assembling full-frame images (FFI)

  • converting EXPMETER_SCI/SKY wavelengths from nm to Angstroms

apply_gain_conversion(chip)

Convert pixel values from ADU to photo-electrons using amplifier-specific gain. Amplifier channels are modified in-place.

Parameters:

chip (str) – CCD identifier, e.g., ‘GREEN’ or ‘RED’.

Return type:

None

Notes

Conversion formula: pixel_electrons = pixel_ADU * gain / 65536

count_amplifiers(chip)

Count the number of amplifier extensions present for a given CCD and determine their channel dimensions.

Parameters:

chip (str) – CCD identifier, e.g., ‘GREEN’ or ‘RED’.

Return type:

None

Notes

Sets instance attributes: - self.namp[chip] : number of amplifier regions detected. - self.dims[chip] : shape of each amplifier channel. Only 2-amp and 4-amp configurations are supported.

info()

Print a summary of the module configuration and processing results.

measure_read_noise(chip, sigma=None, buffer=(5, 5))

Estimate read noise for each amplifier from overscan pixels.

Parameters:

  • chip (str) – CCD identifier, e.g., ‘GREEN’ or ‘RED’.

  • sigma (float, optional) – Threshold for sigma clipping overscan pixels.

  • buffer (tuple of int, optional) – Number of pixels to ignore at the edges (start, end). Defaults to (5, 5).

Return type:

None

Notes

Stores results in: - self.readnoise[channel_ext] : standard deviation of cleaned overscan. - self.rn_nongauss[channel_ext] : non-Gaussian factor computed as std/mad.

orient_channels(chip)

Reorient amplifier channels to a standard orientation in-place. (serial overscan on right, parallel overscan on bottom)

Parameters:

chip (str) – CCD identifier, e.g., ‘GREEN’ or ‘RED’.

Return type:

None

Notes

The transformations are flips, so each is its own inverse — calling twice restores the input. measure_read_noise and subtract_overscan use this to orient to standard, do their work, then restore the original orientation, so a non-standard orientation never propagates between steps; it is also called directly by Quicklook for its L0 display.

static orient_ffi(image, chip)

Flip an assembled image into the standard FFI orientation.

perform(chips=None, *, overscan_method=None, readnoise_sigma=None)

Execute the image assembly algorithm. Optional keyword arguments default to config settings.

Parameters:

  • chips (list of str, optional) – CCD identifiers to process, i.e. ‘GREEN’, ‘RED’

  • overscan_method (str, optional) – Method for overscan subtraction (‘zero’, ‘median’, ‘rowmedian’).

  • readnoise_sigma (float, optional) – Sigma threshold for clipping overscan pixels when measuring read noise.

Returns:

l1_obj – L1 data object containing assembled full frame images (FFIs) for data and variance.

Return type:

KPF1

Notes

Pipeline steps: 1. Count amplifiers and determine dimensions 2. Apply gain conversion (ADU –> electrons) 3. Measure read noise 4. Subtract overscan bias 5. Stitch channels into a full-frame image 6. Convert EXPMETER_SCI/SKY wavelength column labels from nm to Å

Amplifier-channel orientation is handled on-the-fly inside measure_read_noise and subtract_overscan (each restores the original orientation afterward), not as a separate top-level step.

stitch_ffi(chip)

Combine individual amplifier channels into a full-frame image (FFI).

Parameters:

chip (str) – CCD identifier, e.g., ‘GREEN’ or ‘RED’.

Returns:

  • ccd_ffi (ndarray) – Full-frame data image.

  • var_ffi (ndarray) – Full-frame variance image, incorporating read noise.

Notes

Supports 2-amp and 4-amp CCD configurations. Raises an error if any other number of amplifiers is detected.

subtract_overscan(chip, method=None, buffer=(0, 0))

Subtract overscan bias from imaging pixels for each amplifier. Also removes overscan region from amplifier channel, leaving only active imaging area pixels. Amplifier channels are modified in-place.

Parameters:

  • chip (str) – CCD identifier, e.g., ‘GREEN’ or ‘RED’.

  • method (str) – Overscan subtraction method (‘zero’, ‘median’, ‘rowmedian’).

  • buffer (tuple of int, optional) – Number of pixels to ignore at edges. Defaults to (0, 0).

Return type:

None