rawplot

rawplot

Collection of plotting commands to analyze RGB imagers producing Bayer RAW images, such as DSLR cameras or other imagers (i.e RasPi HQ camera) using matplotlib. Its purpose is helping characterize the camera sensor performance and key figures such as gain and read noise.

This is a child project of AZOTEA, an initiative to monitor light pollution through digital imagers. The AZOTEA initiative was started in EU funded project ACTION - Participatory science toolkit against pollution, Grant 824603.

Installation

It is highly recommended to create a virtual environment and activate it, then install rawplot from PyPi

$ mkdir rawplot
$ cd rawplot
$ python3 -m venv .venv
$ source .venv/bin/activate
(.venv) $ pip install rawplot

Usage

The available commands list can be found in the bin virtual environment sub-folder.

ls -1 .venv/bin/raw*
.venv/bin/rawplot-hv
.venv/bin/rawplot-image
.venv/bin/rawplot-imarith
.venv/bin/rawplot-linearity
.venv/bin/rawplot-master
.venv/bin/rawplot-photodiode
.venv/bin/rawplot-plan
.venv/bin/rawplot-ptc
.venv/bin/rawplot-snr
.venv/bin/rawplot-spectral

Common options

All utilities share these command line options, mainly used for debugging purposes:

• ⁻h, --help shows program's options and exit
• --version show program's version numbe and exit.
• --console log debugging messages to console. Default level is INFO.
• --log-file log debugging messages to file. Default level is INFO.
• --verbose raises debugging level to DEBUG.
• --quiet lowers debugging level to WARN.

Other common options using in many commands

subsampling

• --every Picks or subsamples an object retaled with the command (pick every N images in a file list, subsamples points in a histogram by /N, etc.)

Color planes selection

Since we are dealing with RAW images, we are dealing with 4 separate color planes (channels): R Gr Gb B. Most of the commands supports being run on all or a subset of these color planes.

• -c | --channel Specify which of the R, Gr, Gb, or B channels to handle by the command. Can be one or a combination of them. Some commands accept a G channel (an average of Gr and Gb)

Order in the command line is not important. They are internally reordered so that the display & processing order is (R, Gr, Gb, B)

Valid Examples:

--channel B
--channel Gr Gb   
--channel R G B
--channel Gr R B Gb

Invalid Examples:

--channel B B (duplicated)
--channel R A (A is not supported)
--channel R G B Gb (this commbination is not supported)

Region of Interest (ROI)

A ROI is specified given its starting point (x0, y0) and dimensions (width, height). To unifomingly address the different resolutions in camera formats, a normalized ROI is used as input, where both the starting point and dimensions are normalized between 0 and 1. In addition x0+width <= 1.0 and y0+height <= 1.0. When the (x0,y0) starting point is not specified in the command line, the (width,height) dimensions are assumed with respect to the image center.

The different ROI parameters on the command line can be specified as either as decimals or fractions for convenience. Width and height parameters can be specified with -wi | --width and -he | --height.

Example:

--x0 0.2 --y0 0.1 --width 1/3 --height 1/4 

Only when an image file is open, this normalized ROI transforms into a physical ROI with pixels. When the physcal ROI is displayed in several graphs, it is done using NumPy matrix style [y0:y1,x0:x1].

Commands

Examples of usage can be found in chapter 7 (Tools) of GONet all sky camera calibration report along with the dataset on which the usage examples are based.

The camera itself was a Raspberry Pi HQ Camera, installed in a GoNET project device.

Photon Transfer Curves (PTC)

An series of PTC charts, based on the classic Photon Transfer book have been included so far:

CURVE	Description	Units
Curve 1	read, shot, FPN & total noise vs. signal	log rms DN vs. log DN
Curve 1	read, shot, FPN & total noise vs. signal	log rms $e^-$ vs. log $e^-$
Curve 2	read + shot noise vs. signal	log rms DN vs. log DN
Curve 2	read + shot noise vs. signal	log rms $e^-$ vs. log $e^-$
Curve 3	shot noise vs. signal	log rms DN vs. log DN
Curve 3	shot noise vs. signal	log rms $e^-$ vs. log $e^-$
Curve 4	FPN vs. signal	log rms DN vs. log DN
Curve 4	FPN vs. signal	log rms $e^-$ vs. log $e^-$
Curve 5	Read + Shot Noise Variance vs. signal	DN vs. DN
Curve 6	SNR vs. signal	log SNR vs. log DN
Curve 6	SNR vs. signal	log SNR vs. log $e^-$

The curves are based on the following simplified detector model, whose noise variances given by:

\sigma_{TOTAL}^2 = \sigma_{READ}^2 + \sigma_{SHOT}^2 + \sigma_{FPN}^2 \quad [e^-]

Where $\sigma_{READ}$ is the detector read noise, $\sigma_{SHOT}$ is the signal-dependent Poisson noise ($\sigma_{SHOT} = S^{1/2}$), and $\sigma_{FPN}$ is a fixed spatial pattern noise, not all pixels being equal, being also modeled as a signal-dependent noise $\sigma_{FPN} = P_{FPN}S$, where $P_{FPN} \ll 1$.

Working with digital numbers [DN] - instead of electrons - equation above is rewritten as:

\sigma_{TOTAL}^2 = \sigma_{READ}^2 + (S/g) + (P_{FPN}S)^2 \quad [DN]

Miscellaneous Commands

Convenicence commands to help processing an analysis of images using the above commands.

rawplot-master

Utility to make master bias, dark or flat frames from a series of RAW files. Produces a 3D FITS cube, one layer per color.

rawplot-master --console --input-dir images/20240124/biases/ --filter bias* --batch 5 --prefix master --image-type bias

Produces the following result:

2024-02-03 12:38:17,855 [INFO] ============== rawplot.master 0.1.dev89+gc34abee ==============
2024-02-03 12:38:17,855 [INFO] Normalized ROI is [P0=(0.0000,0.0000) DIM=(1.0000 x 1.0000)]
2024-02-03 12:38:17,921 [INFO] The process comprises 8 batches of 5 images max. per batch
2024-02-03 12:38:17,921 [INFO] [1/8] Begin loading 5 images into RAM with R Gr Gb B channels, 2028 x 1520 each
2024-02-03 12:38:18,343 [INFO] [2/8] Begin loading 5 images into RAM with R Gr Gb B channels, 2028 x 1520 each
2024-02-03 12:38:18,727 [INFO] [3/8] Begin loading 5 images into RAM with R Gr Gb B channels, 2028 x 1520 each
2024-02-03 12:38:19,124 [INFO] [4/8] Begin loading 5 images into RAM with R Gr Gb B channels, 2028 x 1520 each
2024-02-03 12:38:19,513 [INFO] [5/8] Begin loading 5 images into RAM with R Gr Gb B channels, 2028 x 1520 each
2024-02-03 12:38:19,892 [INFO] [6/8] Begin loading 5 images into RAM with R Gr Gb B channels, 2028 x 1520 each
2024-02-03 12:38:20,274 [INFO] [7/8] Begin loading 5 images into RAM with R Gr Gb B channels, 2028 x 1520 each
2024-02-03 12:38:20,658 [INFO] [8/8] Begin loading 1 images into RAM with R Gr Gb B channels, 2028 x 1520 each
2024-02-03 12:38:20,794 [INFO] Saving master bias file from 36 images in /home/rafa/repos/own/lica/rawplot/master_bias_frame_aver.fit

rawplot-imarith

Basic image manipulation command. Can be expanded conveniently

rawplot-imarith -h

Arithmetic operations on one or two 3D-FITS cubes

positional arguments:
  {sub}
    sub              Substracts an image or a value (second argument) from a given image (first argument)

options:
  -h, --help         show this help message and exit
  --version          show program's version number and exit
  --console          Log to console.
  --log-file <FILE>  Log to file.
  --verbose          Verbose output.
  --quiet            Quiet output.
  --modules MODULES  comma separated list of modules to activate debug level upon.

rawplot-imarith --console sub master_bias_frame_aver.fit 256

2024-02-07 14:55:43,821 [INFO] ============== rawplot.imarith 0.9.1.dev4+gf89ffa4.d20240207 ==============
2024-02-07 14:55:43,891 [INFO] Created result image on: master_bias_frame_aver_subs.fit

Some example charts produced

Flat field display in all image planes

Histogram of image above

Camera linearity in all channels

Green (Gr) HV Spectrogram

Green (Gr) Photon Transfer Curve

Camera Spectral Response