SIFT features in Computer Vision - Cheat Sheet & Quick Revision

Recall & Review

beginner

What does SIFT stand for in computer vision?

SIFT stands for Scale-Invariant Feature Transform. It is a method to detect and describe local features in images.

beginner

Why is SIFT called 'scale-invariant'?

Because SIFT features can be detected and matched even if the image is resized or zoomed in/out, making them robust to scale changes.

intermediate

What are the main steps in the SIFT algorithm?

1. Detect scale-space extrema using Difference of Gaussians (DoG).<br>2. Localize keypoints and filter out unstable ones.<br>3. Assign orientation to each keypoint.<br>4. Create a descriptor based on local image gradients around the keypoint.

intermediate

How does SIFT handle rotation invariance?

SIFT assigns a dominant orientation to each keypoint based on local gradients, so the descriptor is computed relative to this orientation, making it rotation invariant.

beginner

What is a SIFT descriptor and what is it used for?

A SIFT descriptor is a vector that describes the local image gradients around a keypoint. It is used to match keypoints between different images for tasks like object recognition or image stitching.

What is the first step in the SIFT algorithm?

A Match keypoints between images

B Assign orientation to keypoints

C Create descriptors

D Detect scale-space extrema using Difference of Gaussians

The first step is to detect potential keypoints by finding extrema in the scale-space using Difference of Gaussians.

Why is SIFT robust to changes in image scale?

A It uses color histograms

B It detects features at multiple scales

C It only works on fixed-size images

D It ignores image brightness

SIFT detects features at multiple scales by building a scale-space, making it robust to resizing.

How does SIFT achieve rotation invariance?

A By ignoring orientation

B By resizing the image

C By assigning a dominant orientation to each keypoint

D By using color information

SIFT assigns a dominant orientation to each keypoint and computes descriptors relative to that orientation.

What type of descriptor does SIFT use?

A Histogram of gradients

B Binary descriptor

C Raw pixel values

D Edge detection map

SIFT uses histograms of local image gradients to form its descriptor.

Which of these is NOT a property of SIFT features?

A Color invariance

B Rotation invariance

C Scale invariance

D Robustness to illumination changes

SIFT features are not specifically color invariant; they focus on grayscale gradients.

Explain the main steps of the SIFT algorithm and why each step is important.

Think about how SIFT finds stable points and describes them to match across images.

Describe how SIFT features help in matching objects between two images taken from different distances and angles.

Consider what changes in the images and how SIFT handles them.