fullseye

lf_disparity_to_depth — LIGHTFIELD depth op

使い方

Slope (px per angular step) -> metric depth, the camera-array model.

For a rectified array of viewpoints spaced baseline apart with focal length focal_px expressed in pixels of the sub-aperture image, a point at distance Z shifts by |s| = focal_px * baseline / Z pixels per step. Inverting: Z = focal_px * baseline / |s|, returned in whatever length unit baseline was given in (mm in, mm out).

Only |s| enters, deliberately. The sign of the slope says which side of the focal plane a point is on, and which sign means “nearer” depends on how the angular axis was oriented when the field was decoded — a convention this module cannot know and refuses to guess. If your decode puts near objects at negative slope, negate before calling.

far_depth is the fail-closed switch for the s -> 0 pole (a point at infinity has zero parallax). None (default) makes any |s| < min_slope a ValueError naming how many pixels were affected — mask them, or pass an explicit saturation distance as far_depth and get that value there. There is no silent inf and no silent clamp.

Accepts a scalar or any array; returns the same shape (a scalar in gives a 0-d array out, so downstream code has one type to handle).

Raises ValueError: non-finite slope / focal_px / baseline, a non-positive focal_px / baseline / min_slope, a negative far_depth, and |slope| < min_slope anywhere when far_depth is None.

詳しい使い方ガイド

参考(サンプルデータ・文献)

実行できる例(この op を実際に呼ぶ検証済みサンプル)

型が繋がる次の op(depth を入力に取れる)

同カテゴリ(depth)

lf_depth_from_focus · lf_epi_slope · lf_all_in_focus · lf_plenoptic_design


Provenance: lightfield.py — LIGHTFIELD 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.