SSTV generator in pure Python

PySSTV generates SSTV modulated WAV files from any image that PIL can open (PNG, JPEG, GIF, and many others). These WAV files then can be played by any audio player connected to a shortwave radio for example.

My main motivation was to understand the internals of SSTV in practice, so performance is far from optimal. I tried keeping the code readable, and only performed such optimizations that wouldn't have complicated the codebase.

Command line usage

$ python -m pysstv -h
usage: __main__.py [-h]
              [--mode {MartinM1,MartinM2,ScottieS1,ScottieS2,ScottieDX,Robot36,PasokonP3,PasokonP5,PasokonP7,PD90,PD120,PD160,PD180,PD240,PD290,WraaseSC2120,WraaseSC2180,Robot8BW,Robot24BW}]
              [--rate RATE] [--bits BITS] [--vox] [--fskid FSKID]
              [--chan CHAN] [--resize] [--keep-aspect-ratio]
              [--keep-aspect] [--resample {nearest,bicubic,lanczos}]
              image.png output.wav

Converts an image to an SSTV modulated WAV file.

positional arguments:
  image.png             input image file name
  output.wav            output WAV file name

options:
  -h, --help            show this help message and exit
  --mode {MartinM1,MartinM2,ScottieS1,ScottieS2,ScottieDX,Robot36,PasokonP3,PasokonP5,PasokonP7,PD90,PD120,PD160,PD180,PD240,PD290,WraaseSC2120,WraaseSC2180,Robot8BW,Robot24BW}
                        image mode (default: Martin M1)
  --rate RATE           sampling rate (default: 48000)
  --bits BITS           bits per sample (default: 16)
  --vox                 add VOX tones at the beginning
  --fskid FSKID         add FSKID at the end
  --chan CHAN           number of channels (default: mono)
  --resize              resize the image to the correct size
  --keep-aspect-ratio   keep the original aspect ratio when resizing
                            (and cut off excess pixels)
  --keep-aspect         keep the original aspect ratio when resizing
                            (not cut off excess pixels)
  --resample {nearest,bicubic,lanczos}
                        which resampling filter to use for resizing
                            (see Pillow documentation)

Python interface

The SSTV class in the sstv module implements basic SSTV-related functionality, and the classes of other modules such as grayscale and color extend this. Most instances implement the following methods:

__init__ takes a PIL image, the samples per second, and the bits per sample as a parameter, but doesn't perform any hard calculations
gen_freq_bits generates tuples that describe a sine wave segment with frequency in Hz and duration in ms
gen_values generates samples between -1 and +1, performing sampling according to the samples per second value given during construction
gen_samples generates discrete samples, performing quantization according to the bits per sample value given during construction
write_wav writes the whole image to a Microsoft WAV file

The above methods all build upon those above them, for example write_wav calls gen_samples, while latter calls gen_values, so typically, only the first and the last, maybe the last two should be called directly, the others are just listed here for the sake of completeness and to make the flow easier to understand.