Stereo Rectification Check

easyPRODepth EstimationComputer Vision 1

Verify that pairs of corresponding epipolar points from stereo images lie on the same horizontal scanline.

In a properly rectified stereo pair, all epipolar lines are horizontal, meaning corresponding points in the left and right images should have the same (or very close) y-coordinates. This is a fundamental requirement for efficient stereo matching.

Given two lists of corresponding points from the left and right images, check whether each pair satisfies the horizontal alignment condition within a given tolerance $\epsilon$ :

$|y_{\text{left}} - y_{\text{right}}| \leq \epsilon$

Return a list of boolean values, one per pair, indicating whether the pair is properly rectified.

Example

Input

left_points = [[100, 200], [150, 300], [200, 400]]
right_points = [[80, 200], [130, 301], [180, 410]]
tolerance = 2.0

Output

[True, True, False]

Reasoning

Inputs: left_points and right_points are matched lists of $(x, y)$ pixel coordinates: pair $k$ is the same 3D point seen by the left and the right camera. tolerance is $\epsilon = 2.0$ pixels of allowed vertical misalignment.
Only the $y$ values matter. The $x$ values differ by $20$ in every pair ( $100 \to 80$ , $150 \to 130$ , $200 \to 180$ ); that horizontal shift is disparity, which encodes depth and is expected.
Pair 1: $|200 - 200| = 0 \le 2.0$ , so it is rectified → True.
Pair 2: $|300 - 301| = 1 \le 2.0$ , a one-pixel drift within tolerance → True.
Pair 3: $|400 - 410| = 10 > 2.0$ , the points sit on different scanlines → False.
Output: one boolean per pair, in input order: [True, True, False]. Notice the test is inclusive: a gap of exactly $\epsilon$ still passes, so with $\epsilon = 0$ only identical $y$ values count as rectified.

Constraints

left_points: List of [x, y] coordinates from left image
right_points: List of [x, y] coordinates from right image
tolerance: float (maximum allowed y-coordinate difference)
Return: List of True/False values
Use pure Python

Learn this in

Camera Models & Calibration3D Computer Vision Epipolar Geometry & Stereo Vision3D Computer Vision

More from Computer Vision 1

Semi-Global Matching Cost Aggregationhard Compute Depth Map from Disparitymedium Oriented BRIEF Descriptorhard Gaussian Scale Spacemedium Non-Maximum Suppression (NMS)easy

Stereo Rectification Check

Example

Constraints

Learn this in

More from Computer Vision 1

Background Knowledge

Continue the full explanation

Editor locked