rcc-2d / PixInt2D

This repository contains the renderer-convergence study scripts and PixInt2D, an installable Python package for orthographic 2D rendering of analytic grids and additive speckle textures over one finite-element mesh. PixInt2D supports exact analytic pixel integrals where valid, controlled Rect and Gauss quadrature, robust VTK mapping, Quad9 Newton mapping and a Gaussian camera PSF.

Install

uv pip install -e ".[dev]"

PixInt2D deliberately depends on PyVista/VTK: robust arbitrary single-mesh FE mapping is part of the supported public interface.

Test

pytest -q

The tests use compact committed 32x32 gold renders and do not depend on the large generated out/ directory.

Examples

python -m pixint2d.examples.grid2d_eggbox_rigid
python -m pixint2d.examples.speck2d_disk_rigid
python -m pixint2d.examples.speck2d_gauss_rigid
python -m pixint2d.examples.grid2d_affine
python -m pixint2d.examples.speck2d_affine
python -m pixint2d.examples.grid2d_finite_star
python -m pixint2d.examples.speck2d_finite_star

Each example has inputs as constants at the top of its module and writes a float .npy image plus an 8-bit TIFF preview beneath out/pixint2d_examples/. The finite-star examples use PixInt2D's structured Quad9 Newton inverse map, not the VTK mapping path.

Image convention

For additive speckles, PixInt2D integrates raw overlap coverage first, then clamps/maps it to normalised intensity. Digitisation is explicitly separate:

from pixint2d import quantise_image

codes_u12 = quantise_image(result.image, bits=12)

This preserves one source of truth for all camera bit depths.