• Description This is a python implementation for Fast Texture synthesis using Tree-structured Vector Quantization found here: https://dl.acm.org/doi/pdf/10.1145/344779.345009

• It only handles input images that have dimensions of the form $2^Lm$ where $L$ the levels of the pyramid used and m any integer.
• It only handles grayscale images but the project should be about colored images.
• It hasn't been thoroughly tested.

• Dependencies

• numpy
• matplotlib
• tqdm

Tests

Test1: Test if the output of synth.py and synth.c is equal

1. Generate a fixed “initial_noise.jpg” from Python, which is our Is value
2. The Python synthesizer (synth.py) loads that Is instead of making a new random Is, and run synth.py:

python3 synth.py

3. The translated C version synthesizer (synth.c) loads that same Is, and run synth.c:

gcc -std=c99 -O0 -o synth synth.c -lm
./synth input1.jpg output_c.jpg

4. python3 compare.py output.jpg output_c.jpg

If everything matches, we should see:

MSE: 0.0
Max absolute diff: 0
Wrote difference image to diff.jpg