lensrender op• 資料種類:table → table
• 呼叫: import fullseye as fs; fs.ledger.render_scene_through_lens(system, scene=None, sensor=None, pupil_samples=64, wavelengths=None, exposure_s=0.0001, illuminant=None, seed=0, defocus_mm=0.0, oversample=1, pupil_pattern='halton', pupil_uv=None, pupil_jitter=True, focus_mm=None, add_noise=True, output='float', band_nm=None, source_rings=24) -> 'dict'(要直接呼叫實作,import lensrender; lensrender.render_scene_through_lens(system, scene=None, sensor=None, pupil_samples=64, wavelengths=None, exposure_s=0.0001, illuminant=None, seed=0, defocus_mm=0.0, oversample=1, pupil_pattern='halton', pupil_uv=None, pupil_jitter=True, focus_mm=None, add_noise=True, output='float', band_nm=None, source_rings=24) -> 'dict';從台帳取用則 opsoptics.get("render_scene_through_lens"))
• GPU:無(僅 CPU 參考實作)
透過處方 *system* 的鏡頭拍攝 lens_scene:從感測器一側的光線追跡 -> 字典。
> 以下的詳細說明為原文 —— 摘要與標題已翻譯。
For every pixel (and oversample x oversample points inside it) and every wavelength, the exit pupil is sampled at pupil_samples points (Halton or hexapolar, each pixel's pattern turned by a fixed golden-ratio angle when *pupil_jitter* is on, so the noiseless image does not depend on *seed*), each ray is traced backwards through the lens (trace_sensor_rays), shaded in the scene (scene_radiance) and weighted by cos^2(theta) / d^2 times the pupil-sample area: the Monte Carlo estimate of E = integral L cos(theta) d(omega) over the exit pupil (Kolb, Mitchell & Hanrahan 1995), times (n_image / n_object)^2. Point sources of the scene are added by forward ray tracing. The spectral irradiance becomes electrons per wavelength, E dlambda A_pixel t (lambda / hc) QE(lambda), and the sum goes through the shot noise, the dark electrons, the read noise and the quantiser of lensimage._sensor (*add_noise* False returns the expected values). *sensor* is a dict {"pixel_pitch_um", "size": (rows, cols), "qe", "full_well", "read_noise_e", "dark_e", "bits", "center_mm"} (center_mm moves the window to another point of a larger sensor); *wavelengths* default to the F, d, C lines (each band is as wide as the spacing of the list, or *band_nm*); *illuminant* is the relative spectral shape of the scene's lights (pairs (nm, value)); *defocus_mm* moves the sensor along the light (positive = away from the lens) and *focus_mm* puts it at the paraxial image of an object at that distance (default: the system's own image plane); *oversample* subdivides each pixel. The picture is upright (see the module doc). The table has image (float in 0..1, or uint16 DN with output="uint16"), image_noiseless, channels / channels_noiseless (one image per wavelength, own noise), electrons_mean and electrons_per_wavelength, irradiance_W_m2_nm (K, H, W), pupil_transmission (H, W: the share of the pupil that is not vignetted), the wavelengths and the geometry. Geometric optics only: no diffraction.
optics 的每個運算子都先檢驗輸入再計算(不讓任何東西無聲通過):
• 單位寫進參數名 —— _mm / _um / _deg / _mrad。把 mm 和 µm 弄混不會當掉,而是給出「看似合理卻是錯的答案」,所以用命名來防。這裡絕不從數值大小去猜單位。
• 字串一律 ValueError —— float('50') 會成功,於是未解析的設定值會被當成長度混進來(實測:thin_lens('50', '200') 曾回傳看似合理的 66.667 mm)。bool 也按 True == 1 的隱式提升拒絕。
• complex / masked array 一律 ValueError(僅接受實數槽位;拒絕無聲丟棄虛部或剝掉遮罩)。所有輸入中的 NaN/Inf 一律 ValueError。但已宣告的缺測除外:bottle_fit_from_pattern 的 u_map 以 NaN 接收未能測得的像素(∞ 仍被拒絕)。
• 逐項點名拒絕除零及其近親:焦距 0、曲率半徑 0、折射率 <= 0、全不透明光闌(全為 0,正規化變成 0/0)、總和 <= 0 的 PSF、S0 = 0 的 Stokes 向量、物體位於前焦點(像在無窮遠)。
• 只有兩個運算子會回傳非有限值,而且都寫進了契約:depth_of_field 在超焦距以外回傳 far_mm = inf(這正是超焦距的定義),gaussian_beam 在束腰處回傳 wavefront_radius_mm = inf(平面波前的曲率半徑)。兩者都同時回傳一個有限的夥伴(far_is_infinite / curvature_per_mm)。除此之外的無聲 NaN/Inf 都在內部檢出並 ValueError ——「float64 溢位了」和「答案是無窮大」是兩種不同的主張,不能拿後者的臉去交付前者。但已宣告的缺測除外:bottle_background_map 對光線不可用的像素(全反射等)在 u_mm 中回傳 NaN,並同時回傳 valid 遮罩和數量。
• 尺寸上限:生成網格受 optics.MAX_GRID(4096)限制,傳入的場/PSF/光闌受 optics.MAX_FIELD_ELEMENTS(2^24),ABCD 元件序列受 optics.MAX_SYSTEM_ELEMENTS(1024),Zernike 受 MAX_ZERNIKE_TERMS(512)/ MAX_ZERNIKE_ORDER(40)/ MAX_ZERNIKE_BASIS(2^25)。以 fail-closed 堵住「小參數引發巨大內部配置」的路徑(實測:n_max=40 × 4096² 需要 108 GB)。
• 物理上不可能的狀態同樣拒絕:偏振度 > 1 的 Stokes 向量、負穿透率、負強度、n-|m| 為奇數等非法 Zernike 指標。
• 範例資料目錄(下載 URL / 授權) —— 2-D 用 skimage.data(BSD/公有領域)加合成圖,3-D 給出真實資料源(Stanford/PDS 等)的下載 URL。
• 運算子來歷與參考文獻 —— 該運算子族所依據的研究/方法出處。
• 演算法的正典(作者、年份)與用途見上面的族使用指南。
• poc_photograph_through_prescription — py -3.11 examples/poc_photograph_through_prescription.py
table 作為輸入)abcd_matrix · wavefront_stats · sfr_from_edge · paraxial_trace · seidel_coefficients · spot_stats · tolerance_analysis · wavefront_from_opd
lensrender)lens_scene · exit_pupil_samples · trace_sensor_rays · scene_radiance · bokeh_shape · render_ghost_overlay · render_compare_prescriptions
*Provenance: lensrender.py — OPTICS 運算子登記表。本條目由 tools/opdocs.py md 自動產生(請勿手動編輯)。*
© 2026 Kazufumi Furuse — Fullseye operator documentation. Licensed under Apache-2.0.