Image features

Question 1

Give the definition of feature.

Answer 1

A feature is a “meaningful” part of the image.

Features have two main components
– Feature detection: finding a “stable” (easily detectable) point
– Feature description: a description of the surrounding area

Question 2

What is a feature matching? How can be useful?

Answer 2

1. Evaluate features in two images
2. Find similar features (good matches)

• Object localization
• Stitching image mosaics
• Scene reconstruction
• Place recognition
• Object detection
• Follow patterns in video flows

Question 3

How Harris corners are computed?

Answer 3

• Consider a patch in a given position (𝑥𝑖, 𝑦𝑖)
• Consider a displacement Δ𝑥, Δ𝑦
• Similarity is measured by means of the autocorrelation, a function of the displacement

Question 4

Consider Harris corners detector: describe the auto-correlation matrix.

Answer 4

Matrix

Studying the eigenvalues we get information about the type of patch.

Question 5

Describe SUSAN corner detector.

Answer 5

– Analyzes a circular window around the point
– No derivatives involved
– Edge+corner detector
– Robust to noise

Question 6

What is a blob?

Answer 6

Feature!

A blob is a region where
– Considered properties are different from surrounding regions
– Properties are (approximately) constant inside the region

Question 7

What is the Maximally Stable Extremal Regions (MSER)?

Answer 7

Algorithm that uses blobs!

1. Apply a series of thresholds (e.g., one for each gray level)
2. Compute the connected binary regions
3. Compute some statistics for each region
   • E.g.: area, convexity, circularity, …
4. Analyze how persisten each blob is

Question 8

How can we create different scales of an image?

Answer 8

Using the N-dimensional gaussian kernel and varying t.

Give the formula!

Question 9

What is the SIFT and what are its strong points?

Answer 9

• Very reliable keypoint detector and descriptor.
• Image content mapped into local feature
  coordinates.

Strong points:
– Local – robust to occlusions
– Distinctive – distinguish objects in large databases
– Dense – many features can be found even on small objects
– Efficiency – fast computation

Question 10

Describe the SIFT algorithm.

Answer 10

1. Scale-space extrema (no local) detection: use Difference of Gaussians (DoG)
2. Keypoint localization and refinement: search for maxima and minima of the DoG and refine them using the Hessian matrix (2nd order derivative)
3. Orientation measurement: compute gradient magnitude and orientation in the keypoint neighborhood (not with the Sobel operators) - takes also the 2nd peak for duplicating the keypoint
4. Descriptor calculation: create histograms and the descriptor vector

Question 11

What are the Haar features for face detection?

Answer 11

Rectangular filters.

Local feature: subtract the sum of pixels in the white areas from the sum of pixels in the black area.

Question 12

Viola and Jones algortihm.

Answer 12

1. Define Haar features
2. Weak learners working on Haar features
3. Boosting for combining multiple weak learners
4. Cascade of classifiers

Question 13

Other SIFT-based features.

Answer 13

• PCA-SIFT
• SURF (Speeded Up Robust Features)
• GLOH (Gradient Location-Orientation Histogram)