Computer Vision 2 Marks Short Answers

1. Define Computer Vision:

Computer Vision is a branch of AI that enables computers to interpret and understand visual data,

such as images and videos, through algorithms. It involves tasks like object detection, image

classification, and 3D reconstruction. Applications include autonomous vehicles, facial recognition,

and medical diagnostics.

2. List the different color spaces in image processing:

Common color spaces are:

- RGB: Red, Green, Blue used in screens.

- HSV: Hue, Saturation, Value for color-based segmentation.

- YCbCr: Luminance and chrominance for video processing.

- LAB: Perceptual uniformity for advanced image analysis.

3. State GrabCut algorithm:

GrabCut is a segmentation algorithm that uses a graph-based approach to separate the foreground

and background. It requires an initial bounding box or user input to identify regions of interest. The

algorithm iteratively optimizes segmentation using energy minimization. It is widely used in image

editing tools.

4. List the feature detectors and feature descriptors:

- Feature Detectors: SIFT, SURF, FAST, Harris.

- Feature Descriptors: ORB, BRIEF, FREAK.

Detectors find keypoints in an image, while descriptors encode their characteristics for matching.

5. What is Multi-frame structure from motion?

It is a method to reconstruct 3D structures and motion of objects using multiple 2D images taken

from different viewpoints. The approach estimates depth and motion parameters. It is widely used in
robotics and augmented reality.

6. Define hierarchical motion alignment:

Hierarchical motion alignment aligns motion data in a layered manner, starting with coarse-level

adjustments and refining finer details. This approach helps manage large-scale motion and optimize

accuracy. It is used in video stabilization and motion tracking.

7. List the 3D reconstruction techniques in computer vision:

- Stereo Vision: Using two images for depth estimation.

- Structure from Motion (SfM): Reconstructs 3D structure from multiple images.

- Photometric Stereo: Uses lighting variations to infer depth.

- Depth Sensors: Devices like LiDAR or Kinect.

8. What are the various methods to infer shape?

- Shape from Shading: Uses light intensity.

- Shape from Silhouette: Uses object outlines.

- Shape from Texture: Analyzes texture distortions.

- Shape from Motion: Observes motion patterns.

9. List the two applications of image recognition:

- Face Recognition: Used in security systems and smartphones.

- Object Recognition: Identifying objects for robotics and retail systems.

10. Define the term layers and sprites:

- Layers: Represent data or processing stages in image editing or neural networks.

- Sprites: 2D graphical objects used in animations and video games.

11. Define histogram equalization:

A technique to enhance image contrast by redistributing pixel intensity values. It stretches the range

of brightness to make details more visible. Commonly used in medical imaging and photography.
12. Define computer vision and state its applications:

Computer Vision enables machines to analyze and interpret visual inputs. Applications include

medical imaging, industrial inspection, autonomous vehicles, and surveillance systems.

13. Name the feature detectors and feature descriptors:

- Detectors: SIFT, SURF, FAST, Harris.

- Descriptors: ORB, BRIEF, FREAK.

These tools extract and describe keypoints for tasks like matching and tracking.

14. Illustrate segmentation:

Segmentation divides an image into regions or objects. For example, separating a person from the

background in a photograph. Techniques include thresholding, edge detection, and clustering.

15. What is factorization?

Factorization decomposes a matrix into simpler components. In computer vision, it is used for

motion analysis to estimate the 3D structure of a scene from 2D projections.

16. Define bundle adjustment:

Bundle adjustment optimizes 3D points and camera parameters to minimize reprojection error in

multi-view geometry. It refines initial 3D reconstruction results.

17. Define RANSAC algorithm:

RANSAC (Random Sample Consensus) estimates model parameters by iteratively selecting

random subsets of data. It is robust against noise and outliers, commonly used in line fitting and

homography estimation.

18. Define shape from X:

Shape from X refers to methods that infer 3D shapes from visual cues like shading, motion, or

texture. These techniques are useful in computer vision for 3D modeling.

Environment mattes capture how objects interact with their surroundings in terms of reflection and

transparency. This is essential for realistic rendering in graphics and video compositing.

20. Summarize instance recognition:

Instance recognition identifies specific objects or instances within a category, such as recognizing a

particular car brand. It is used in object tracking, retail, and autonomous navigation.