"""
Simple reference subtraction utilities.
Functions
---------
ref_subtraction_channel
Channel-based reference subtraction.
ref_subtraction_row
Row-based reference subtraction.
"""
import numpy as np
[docs]
def ref_subtraction_channel(image, channel_start=0, channel_end=128, use_ref_channel=False):
"""
Channel-based reference subtraction.
Performs a simple channel-wise reference pixel subtraction on the slopes image.
Calculates a linear fit to the median pixel values at the top and bottom of each channel,
and subtracts the fitted line from each column in the channel.
The image is updated in place; the updated image is also the return value.
Parameters
----------
image : np.array
a 2D numpy array representing the slopes image.
channel_start : int, optional
The starting index for the first channel.
channel_end : int, optional
The ending index for the first channel.
use_ref_channel : bool, optional
Whether to use the reference output.
Returns
-------
image : np.array
A 2D numpy array with the reference pixel values subtracted from each column in each channel.
The image is expected to have 33 channels, each with 128 columns.
"""
# Define beginning and ending indices for the first channel as an initial starting point
n_channels = 32
if use_ref_channel:
n_channels = n_channels + 1
# Vertical reference pixel subtraction
for _ in range(0, n_channels):
ch = image[:, channel_start:channel_end]
bottom_med = np.median(ch[0:4, :])
top_med = np.median(ch[4092:4096, :])
# Do a least squares fit solution to fit a line to the top and bottom medians
points = [(1.5, bottom_med), (4093.5, top_med)]
x_coords, y_coords = zip(*points, strict=False)
A = np.vstack([x_coords, np.ones(len(x_coords))]).T
m_cor, c_cor = np.linalg.lstsq(A, y_coords, rcond=None)[0]
# For all the columns, compute the value from the lstsq fit, and subtract it from all
# pixels in that col
for j in range(0, 4096):
# Use the lstsq fit from above to compute a reference pixel median value to be subtracted
# from each col
I_el = m_cor * j + c_cor
image[j, channel_start:channel_end] = image[j, channel_start:channel_end] - I_el
# Iterate
channel_start = channel_start + 128
channel_end = channel_end + 128
return image
[docs]
def ref_subtraction_row(image, use_ref_channel=False, slope=None):
"""
Row-based reference subtraction.
Performs a simple row-wise reference pixel subtraction on the slopes image.
Fits active-region median as a funciton of reference-region median, subtracts the
fitted median from each row.
The image is updated in place; the updated image is also the return value.
Parameters
----------
image : np.array
A 2D numpy array representing the slopes image.
use_ref_channel : bool, optional
Whether to use the reference output for fitting.
slope : float, optional
The multiplying factor by the reference to subtract from the data.
If None, then does a fit to determine the best slope.
Returns
-------
image : np.array
A 2D numpy array with the reference pixel values subtracted from each row.
"""
sci_medians = []
ref_medians = []
for row in range(0, 4096):
sci_medians.append(np.median(image[row, 4:4092]))
if use_ref_channel:
ref_medians.append(np.median(image[row, 4096:4224]))
else:
ref_medians.append(np.median(np.hstack((image[row, 0:4], image[row, 4092:4096]))))
ref_medians = np.array(ref_medians)
m_med, _ = np.polyfit(ref_medians, sci_medians, 1)
ctr = np.median(ref_medians)
# overwrite with provided slope if specified
if slope is not None:
m_med = slope
# Iterate through all rows: compute median value, subtract it across the whole row
for i in range(0, 4096):
image[i, :] = image[i, :] - m_med * (ref_medians[i] - ctr)
return image