Skip to content

Quiz: Image Processing and Computer Vision

Test your understanding of image representation, filtering, and computer vision concepts.

1. A grayscale image is represented as:

  1. A 3D tensor with RGB channels
  2. A 2D matrix of intensity values
  3. A 1D vector of pixel positions
  4. A list of color names
Show Answer

The correct answer is B. A grayscale image is a 2D matrix where each entry represents the intensity (brightness) of a pixel, typically ranging from 0 (black) to 255 (white).

Concept Tested: Grayscale Image

2. An RGB image is stored as:

  1. A single 2D matrix
  2. Three separate matrices (one per color channel)
  3. A 1D array of values
  4. A text file
Show Answer

The correct answer is B. An RGB image consists of three 2D matrices (or a 3D tensor), one for each color channel: Red, Green, and Blue. Each pixel has three values.

Concept Tested: RGB Image

3. Image convolution:

  1. Increases image resolution
  2. Applies a kernel to extract features or transform the image
  3. Converts color to grayscale
  4. Compresses the image
Show Answer

The correct answer is B. Convolution slides a kernel (small matrix) across the image, computing weighted sums at each position. Different kernels produce different effects like blurring, sharpening, or edge detection.

Concept Tested: Image Convolution

4. A Gaussian blur filter:

  1. Sharpens edges in the image
  2. Smooths the image by averaging with Gaussian-weighted neighbors
  3. Detects vertical edges only
  4. Increases image contrast
Show Answer

The correct answer is B. Gaussian blur uses weights that follow a Gaussian distribution, giving more weight to nearby pixels. This produces smooth, natural-looking blur without blocky artifacts.

Concept Tested: Blur Filter

5. The Sobel operator is used for:

  1. Image compression
  2. Edge detection by computing image gradients
  3. Color space conversion
  4. Image resizing
Show Answer

The correct answer is B. The Sobel operator approximates image derivatives (gradients) to detect edges. It uses two kernels: one for horizontal gradients (\(G_x\)) and one for vertical gradients (\(G_y\)).

Concept Tested: Sobel Operator

6. The Fourier transform of an image reveals:

  1. The color distribution
  2. Frequency components (how quickly intensity changes)
  3. The number of objects
  4. The image dimensions
Show Answer

The correct answer is B. The Fourier transform decomposes an image into frequency components. Low frequencies represent gradual changes (smooth areas); high frequencies represent rapid changes (edges, textures).

Concept Tested: Fourier Transform

7. SVD-based image compression works by:

  1. Removing all color information
  2. Keeping only the largest singular values and discarding the rest
  3. Reducing the number of pixels
  4. Converting to a different file format
Show Answer

The correct answer is B. SVD compression keeps only the \(k\) largest singular values, which capture the most important structure. The approximation \(A_k = \sum_{i=1}^k \sigma_i u_i v_i^T\) requires less storage.

Concept Tested: Image Compression

8. In Harris corner detection, a corner is characterized by:

  1. Only one large eigenvalue of the structure tensor
  2. Two large eigenvalues of the structure tensor
  3. Zero eigenvalues
  4. Negative eigenvalues
Show Answer

The correct answer is B. At a corner, the structure tensor has two large eigenvalues, indicating significant intensity changes in both directions. An edge has one large eigenvalue; flat regions have none.

Concept Tested: Feature Detection

9. A homography transformation:

  1. Only handles rotation
  2. Maps points between two planes, handling perspective
  3. Is always the identity
  4. Only works on 1D signals
Show Answer

The correct answer is B. A homography is a 3×3 projective transformation that maps points between two planes. It can represent perspective distortion, rotation, scaling, and translation.

Concept Tested: Homography

10. The convolution theorem states that convolution in spatial domain equals:

  1. Addition in frequency domain
  2. Multiplication in frequency domain
  3. Division in frequency domain
  4. Convolution in frequency domain
Show Answer

The correct answer is B. The convolution theorem states that \(f * g = \mathcal{F}^{-1}(\mathcal{F}(f) \cdot \mathcal{F}(g))\). Convolution in spatial domain corresponds to element-wise multiplication in frequency domain.

Concept Tested: Frequency Domain