Feature and Corner Detection Practice Flashcards

Comprehensive fill-in-the-blank flashcards covering the key concepts of feature detection, the Moravec detector, and the Harris corner detector based on Lecture 7 and 8 notes.

Feature detection mainly aims to extract meaningful __________ information from images.

structural

An example of a point feature is a(n) __________.

Corner

Edges in an image correspond to strong __________ transitions.

intensity

The application of fingerprint recognition relies heavily on __________ detection.

minutiae

In a flat region, shifting a window in any direction results in __________ SSD values.

small

In an edge region, intensity variation occurs in __________ only.

one direction

The Moravec detector identifies corners by computing the __________.

Sum of squared differences (SSD)

Under the Moravec detector, a pixel is classified as a corner when the __________ across shifts is large.

minimum SSD

Two main limitations of the Moravec detector are sensitivity to noise and a lack of __________ invariance.

rotation

The Harris detector uses __________ to analyze the local image structure.

image gradients

The structure tensor matrix $M$ is constructed using the terms $I_x^2$, $I_y^2$, and __________.

$$I_x I_y$$

In the structure tensor matrix, the trace represents the __________.

total intensity variation

If both eigenvalues of the structure tensor matrix are large, the pixel corresponds to a(n) __________.

corner

The Harris response function is defined by the expression $R = \_\_\_\_\_\_\_\_\_\_$.

$$det(M) - \alpha \, trace(M)^2$$

Typical values for the parameter $\alpha$ in Harris detection lie in the range __________.

$$0.04 - 0.06$$

If the Harris response value $R$ is negative with a large magnitude, the region is classified as a(n) __________.

edge

Horizontal gradients ($I_x$) are often computed using the differentiation filter __________.

$$[-1, 0, 1]$$

When the parameter $\alpha$ is chosen too large, true corners might be incorrectly classified as __________.

edges

Distinctive image locations that can be reliably detected and matched across multiple images are called __________.

interest points

Eye corners, the nose tip, and mouth boundaries are examples of __________.

facial landmark features

The SSD function used in Moravec detection is mathematically expressed as $E_{m,n}(x, y) = \_\_\_\_\_\_\_\_\_\_$.

$$\sum_{(u,v) \in W} [I(m + u, n + v) - I(m + x + u, n + y + v)]^2$$

The structure tensor matrix is defined as $M = \_\_\_\_\_\_\_\_\_\_$.

$\begin{pmatrix} \sum I_x^2 & \sum I_x I_y \\ \sum I_x I_y & \sum I_y^2 \end{pmatrix}$. (Note: In the transcript, I2x refers to $I_x^2$)

Increasing the __________ in Harris detection reduces the number of detected points by selecting only pixels with very large response values.

threshold value

In a scatter plot of gradient values $(I_x, I_y)$, a corner is indicated when gradient distributions spread in __________.

two directions