Computer Vision Lecture Notes

These flashcards cover key concepts and definitions from the lecture on 3D computer vision, focusing on stereo vision and related geometries.

What is the minimum requirement for depth estimation in image-based 3D reconstruction?

At least two cameras are required.

What does the essential matrix represent in stereo vision?

It represents the geometric relationship between two calibrated images.

What is the significance of disparity in standard stereo vision?

Disparity is the difference in point location between images, used for depth estimation.

What is epipolar geometry?

It describes the geometric relationship between two views of a scene, where corresponding points lie on specific lines called epipolar lines.

Why is the rectification of stereo images important?

It allows corresponding pixels to be aligned in the same row for improved stereo matching.

What is the primary challenge in correspondence-based stereo vision?

Pattern matching in images can be challenging due to variations in light and occlusions.

How many point correspondences are needed to estimate the fundamental matrix using the eight-point method?

At least 8 point correspondences are typically required.

What can cause ambiguity in depth estimation when using shape from shading techniques?

Intensity changes can lead to more than one valid interpretation of surface orientation.

What are the main components of the fundamental matrix estimation?

The fundamental matrix estimation involves correspondence between points, linear constraints, and normalization.

What are the benefits of using wide baseline stereo cameras?

They allow for larger disparities and more accurate depth estimation.

What is the minimum requirement for depth estimation in image-based 3D reconstruction?

At least two cameras are required.

What does the essential matrix represent in stereo vision?

It represents the geometric relationship between two calibrated images.

What is the significance of disparity in standard stereo vision?

Disparity is the difference in point location between images, used for depth estimation.

What is epipolar geometry?

It describes the geometric relationship between two views of a scene, where corresponding points lie on specific lines called epipolar lines.

Why is the rectification of stereo images important?

It allows corresponding pixels to be aligned in the same row for improved stereo matching.

What is the primary challenge in correspondence-based stereo vision?

Pattern matching in images can be challenging due to variations in light and occlusions.

How many point correspondences are needed to estimate the fundamental matrix using the eight-point method?

At least 8 point correspondences are typically required.

What can cause ambiguity in depth estimation when using shape from shading techniques?

Intensity changes can lead to more than one valid interpretation of surface orientation.

What are the main components of the fundamental matrix estimation?

The fundamental matrix estimation involves correspondence between points, linear constraints, and normalization.

What are the benefits of using wide baseline stereo cameras?

They allow for larger disparities and more accurate depth estimation.

What is the process of triangulation in 3D reconstruction?

Triangulation is the method of determining the 3D position of a point from its 2D projections in two or more images, using the camera parameters.

What is the purpose of camera calibration in 3D reconstruction?

Camera calibration determines the intrinsic (focal length, principal point, distortion) and extrinsic (rotation, translation) parameters of a camera, essential for accurate 3D measurements.

What distinguishes the fundamental matrix from the essential matrix?

The essential matrix is used for calibrated cameras, while the fundamental matrix applies to uncalibrated cameras, representing the epipolar geometry without requiring intrinsic parameters.

Name two general categories of stereo matching algorithms.

Stereo matching algorithms are generally categorized into local (e.g., window-based) and global (e.g., graph cuts, dynamic programming) approaches.