imaging_sim opimage2d × table → image2dimport fullseye as fs; fs.ledger.render_through_lens(image, system, pixel_pitch_um=5.5, field_of_view=None, zones=3, noise=None, seed=0, illumination='traced', size=None, oversample=None) (実装を直接呼ぶなら import lensimage; lensimage.render_through_lens(image, system, pixel_pitch_um=5.5, field_of_view=None, zones=3, noise=None, seed=0, illumination='traced', size=None, oversample=None)、台帳から引くなら opsoptics.get("render_through_lens"))Render an ideal irradiance image as the sensor behind system would record it (image2d).
image (H×W, non-negative) is the ideal (paraxial) image on a sensor of
pixel_pitch_um pixels centred on the optical axis; with field_of_view
(half field to the sensor corner: degrees for an object at infinity,
object height in mm otherwise) the picture is first zoomed so the corner
sees that field. Pipeline: (a) inverse distortion remap
(:func:distortion_map grid, scipy.ndimage.map_coordinates order 1);
(b) spatially varying blur — a zones×zones lattice of tile centres,
each with its own pixel-integrated :func:psf_from_opd (the +y-field PSF
rotated to the tile azimuth), blended with bilinear (tent) weights so
seams vanish; (c) relative illumination: "traced" = fraction of the
field’s ray bundle that reaches the image (vignetting, from
:func:raytrace.ray_bundle) normalised to the axis, times cos⁴ (obliquity,
objects at infinity only), "cos4" = the classic law alone, "none";
(d) sensor, when noise is True or a dict {"full_well": 20000,
"read_e": 3.0, "bits": 12, "exposure": 1.0, "dark_e": 0.0}: electrons =
irradiance × exposure × full_well, Poisson shot noise
(:func:photoncount.photon_sample), Gaussian read noise, quantisation to
bits, returned as DN/(2^bits − 1). With noise=None the float
irradiance is returned untouched (deterministic; the noisy path is
deterministic for a given seed too).
oversample defaults to whatever keeps at least 2 PSF samples per pixel
(max(4, ceil(2·λ·F#/pitch))). Ground truth: a δ image through the
f/2 paraboloid gives the pixel-integrated Airy PSF; a checkerboard through
it comes back undistorted (correlation > 0.99); energy is conserved to 1 %
with illumination off; noise off is bit-reproducible.
optics の全 op は入力を検証してから計算する(黙って通さない):
_mm / _um / _deg / _mrad。mm と µm の取り違えは crash ではなく「もっともらしく間違った答え」なので、名前で防ぐ。大きさから単位を推測する処理は一切しない。ValueError — float('50') は成功してしまうため、未パースの設定値が長さとして通り抜ける(実測: thin_lens('50', '200') がもっともらしい 66.667 mm を返していた)。bool も True == 1 の暗黙昇格として拒否。ValueError(実数枠のみ。虚部の無言切り捨て・マスク剥がしを拒否)。NaN/Inf は全入力で ValueError。depth_of_field の過焦点距離以遠の far_mm = inf(それが過焦点距離の定義)と gaussian_beam のウエストでの wavefront_radius_mm = inf(平面波面の曲率半径)。どちらも有限の相棒(far_is_infinite / curvature_per_mm)を併せて返す。それ以外の無言 NaN/Inf は内部で検出して ValueError —「float64 が溢れた」と「答えが無限大」は別の主張なので、後者の顔で前者を返さない。optics.MAX_GRID(4096)、供給された場/PSF/開口は optics.MAX_FIELD_ELEMENTS(2^24)、ABCD 素子列は optics.MAX_SYSTEM_ELEMENTS(1024)、Zernike は MAX_ZERNIKE_TERMS(512)/ MAX_ZERNIKE_ORDER(40)/ MAX_ZERNIKE_BASIS(2^25)。小さな引数から巨大な内部確保が起きる経路(実測: n_max=40 × 4096² で 108 GB)を fail-closed で塞ぐ。| 物理的に不可能な状態も拒否: 偏光度 > 1 の Stokes ベクトル、負の透過率、負の強度、n- | m | が奇数などの不正な Zernike 添字。 |
py -3.11 examples/lens_defect_dataset_demo.pyimage2d を入力に取れる)fraunhofer_pattern · pupil_psf · pupil_blur · psf_to_mtf · illumination_uniformity · surface_defect · defocus_blur
imaging_sim)psf_from_opd · distortion_map · defect_dataset · calibration_views
Provenance: lensimage.py — OPTICS operator registry. この per-op ノートは tools/opdocs.py md が自動生成(手編集しない)。
© 2026 Kazufumi Furuse — Fullseye operator documentation. Licensed under Apache-2.0.