wayground logo

Free Printable Worksheets

Font size

S
M
L
XL
Worksheets

ITA0511 – Computer Vision for Image Processing: Class Test 3

Total questions: 100

Worksheet time: 50mins

Name
Class
Date
1.

A 512×512 image is downsampled to 256×256 without anti-aliasing. The most likely visual artifact is:

a)

Blurring

b)

Aliasing

c)

Quantization noise

d)

Contrast loss

2.

If spatial resolution is doubled in both x and y directions, total number of pixels increases by:

a)

b)

c)

d)

3.

Reducing intensity levels from 256 to 16 primarily increases:

a)

Spatial noise

b)

Quantization error

c)

Blur

d)

Edge thickness

4.

A 10-bit grayscale image has how many intensity levels?

a)

256

b)

512

c)

1024

d)

2048

5.

Which operation improves intensity resolution but not spatial resolution?

a)

Sampling

b)

Quantization

c)

Interpolation

d)

Smoothing

6.

For highlighting low-intensity details in an image, the best transformation is:

a)

Histogram equalization

b)

Log transformation

c)

High-pass filtering

d)

Negative transformation

7.

Histogram equalization may fail when:

a)

Image has uniform histogram

b)

Image has local contrast variations

c)

Image is grayscale

d)

Image is large

8.

Adaptive histogram equalization is preferred because it:

a)

Reduces noise

b)

Works on local regions

c)

Is faster

d)

Preserves edges

9.

Increasing y (>1) in power-law transformation causes the image to:

a)

Brighten

b)

Darken

c)

Blur

d)

Sharpen

10.

Contrast stretching mainly modifies:

a)

Frequency components

b)

Intensity range

c)

Spatial resolution

d)

Image size

11.

A median filter is best for removing:

a)

Gaussian noise

b)

Salt-and-pepper noise

c)

Periodic noise

d)

High-frequency noise

12.

Gaussian smoothing is applied before edge detection to:

a)

Strengthen edges

b)

Reduce noise sensitivity

c)

Increase resolution

d)

Improve contrast

13.

Mean filter causes edge blurring because it:

a)

Is non-linear

b)

Averages neighboring pixels

c)

Amplifies gradients

d)

Detects corners

14.

Laplacian filter enhances edges but is sensitive to noise because it is:

a)

Low-pass

b)

Non-linear

c)

Second-order derivative

d)

Binary

15.

Which filter preserves edges while smoothing flat regions?

a)

Mean

b)

Median

c)

Gaussian

d)

Laplacian

16.

Low-pass filtering in frequency domain removes:

a)

Edges

b)

Noise

c)

Smooth regions

d)

Texture

17.

DCT is preferred in JPEG because it provides:

a)

Lossless compression

b)

Energy compaction

c)

Edge detection

d)

Noise suppression

18.

Removing all high-frequency components results in:

a)

Sharpened image

b)

Blurred image

c)

No change

d)

Binary image

19.

Most image energy is concentrated in:

a)

High frequencies

b)

Mid frequencies

c)

Low frequencies

d)

Random frequencies

20.

Frequency domain enhancement operates on:

a)

Pixels

b)

Neighborhoods

c)

Transform coefficients

d)

Gradients

21.

Edge detection is based on detecting:

a)

Intensity values

b)

Zero crossings

c)

Rapid intensity changes

d)

Color variation

22.

Sobel operator gives stronger edges than Prewitt because it:

a)

Uses larger kernel

b)

Includes smoothing

c)

Uses frequency domain

d)

Is non-linear

23.

Canny detector uses double thresholding to:

a)

Reduce noise

b)

Remove false edges

24.

Laplacian edge detector often produces double edges because it:

a)

Uses gradients

b)

Detects zero crossings

c)

Is non-linear

d)

Uses thresholding

25.

Which edge detector is most robust to noise?

a)

Roberts

b)

Prewitt

c)

Sobel

d)

Canny

26.

Global thresholding fails when the image has:

a)

Uniform background

b)

High contrast

c)

Varying illumination

d)

Binary regions

27.

Region-based segmentation groups pixels based on:

a)

Edge strength

b)

Similarity

c)

Gradient direction

d)

Noise level

28.

Watershed segmentation often causes over-segmentation due to:

a)

Noise and local minima

b)

Poor edges

c)

Low resolution

d)

Thresholding

29.

Edge-based segmentation may fail when:

a)

Objects have smooth boundaries

b)

Edges are weak or broken

c)

Image is binary

d)

Contrast is high

30.

Segmentation is application-dependent because:

a)

Images vary

b)

Objects vary in appearance

c)

No universal method exists

d)

All of the above

31.

Corners are preferred features because they have:

a)

High intensity

b)

High variation in all directions

c)

Uniform gradient

d)

Smooth texture

32.

Harris detector fails on edges because:

a)

Gradients exist in one direction

b)

Noise dominates

c)

Eigenvalues are equal

d)

Threshold is high

33.

SIFT features are invariant to scale due to:

a)

Normalization

b)

Gaussian pyramid

c)

Histogram equalization

d)

Thresholding

34.

PCA reduces dimensionality by selecting components with:

a)

Minimum variance

b)

Maximum variance

c)

Minimum mean

d)

Maximum entropy

35.

Retaining fewer PCA components mainly removes:

a)

Important features

b)

Redundant information

c)

Edges

d)

Noise only

36.

Morphological erosion primarily:

a)

Expands objects

b)

Shrinks objects

c)

Smooths edges

d)

Detects corners

37.

Opening operation is useful for removing:

a)

Large objects

b)

Thin noise

c)

Edges

d)

Texture

38.

Closing operation is used to:

a)

Remove small holes

b)

Detect boundaries

c)

Sharpen edges

d)

Reduce noise

39.

CNNs outperform classical methods because they:

a)

Use fixed filters

b)

Learn hierarchical features

c)

Reduce noise

d)

Use PCA

40.

Pooling layers in CNNs mainly reduce:

a)

Depth

b)

Feature map size

c)

Kernel size

d)

Noise

41.

Image preprocessing improves:

a)

Storage only

b)

Noise only

c)

Feature extraction accuracy

d)

Image size

42.

Feature extraction follows segmentation because:

a)

Regions are defined

b)

Noise is removed

c)

Image is resized

d)

Histogram is flattened

43.

Object recognition accuracy depends heavily on:

a)

Camera resolution

b)

Feature quality

c)

File format

d)

Image size

44.

Real-time vision systems prefer SURF over SIFT due to:

a)

Accuracy

b)

Robustness

c)

Speed

d)

Memory

45.

PCA before classification mainly reduces:

a)

Training time

b)

Image contrast

c)

Noise only

d)

Pixel count

46.

A 256×256 image divided into 8×8 blocks produces:

a)

512 blocks

b)

1024 blocks

c)

2048 blocks

d)

4096 blocks

47.

Quantizing 256 levels into 16 levels gives a step size of:

a)

8

b)

16

c)

32

d)

64

48.

If PCA retains 95% variance with 40 components, reducing to 20 components will most likely:

a)

Increase accuracy

b)

Slightly reduce accuracy

c)

Improve contrast

d)

Increase noise

49.

Removing low-frequency components affects the image by:

a)

Removing edges

b)

Removing smooth regions

c)

Increasing brightness

d)

Reducing noise

50.

The first step in a computer vision pipeline is:

a)

Segmentation

b)

Feature extraction

c)

Image acquisition

d)

Classification

51.

A digital image is represented as:

a)

Continuous signal

b)

Discrete intensity values

c)

Binary waveform

d)

Analog voltage

52.

Image preprocessing mainly aims to:

a)

Reduce file size

b)

Improve image quality

c)

Classify objects

d)

Detect feature

53.

Noise in images generally affects:

a)

Low frequencies

b)

High frequencies

c)

Color depth

d)

Image resolution

54.

Salt-and-pepper noise is best removed using:

a)

Mean filter

b)

Median filter

c)

Gaussian filter

d)

Laplacian filter

55.

Gaussian smoothing is mainly used to:

a)

Enhance edges

b)

Remove noise

c)

Detect corners

d)

Segment images

56.

Histogram equalization improves:

a)

Brightness

b)

Contrast

c)

Resolution

d)

Noise

57.

Contrast stretching is a:

a)

Frequency domain technique

b)

Point processing technique

c)

Segmentation method

d)

Compression technique

58.

Negative image transformation is useful for:

a)

Medical images

b)

Satellite images

c)

Low contrast images

d)

All of the above

59.

Adaptive histogram equalization works on:

a)

Entire image

b)

Fixed window size

c)

Local regions

d)

Edge pixels

60.

Which transformation highlights dark regions?

a)

Log transformation

b)

Power-law (γ < 1)

c)

Histogram equalization

d)

High-pass filter

61.

Mean filter is mainly used for:

a)

Edge detection

b)

Noise smoothing

c)

Image sharpening

d)

Feature extraction

62.

High-pass filtering emphasizes:

a)

Smooth regions

b)

Low frequencies

c)

Edges and details

d)

Background

63.

Laplacian filter is a:

a)

First-order derivative

b)

Second-order derivative

c)

Zero-order filter

d)

Non-linear filter

64.

Which filter preserves edges while smoothing?

a)

Mean filter

b)

Median filter

c)

Gaussian filter

d)

Box filter

65.

Sobel operator computes:

a)

Laplacian

b)

Gradient magnitude

c)

Histogram

d)

Frequency response

66.

Fourier Transform converts image from:

a)

Spatial to frequency domain

b)

Frequency to spatial domain

c)

Pixel to binary

d)

RGB to grayscale

67.

Low-pass filtering removes:

a)

Smooth regions

b)

Noise

c)

Edges

d)

Texture

68.

High-pass filters are used for:

a)

Blurring

b)

Smoothing

c)

Sharpening

d)

Compression

69.

DCT is widely used in:

a)

PNG

b)

JPEG

c)

BMP

d)

TIFF

70.

Frequency domain enhancement operates on:

a)

Pixels

b)

Gradients

c)

Coefficients

d)

Neighborhoods

71.

Edge detection finds:

a)

Object interiors

b)

Object boundaries

c)

Color variation

d)

Texture patterns

72.

Canny edge detector is preferred because it:

a)

Is fast

b)

Reduces noise and detects edges accurately

c)

Uses simple masks

d)

Needs no thresholds

73.

Thresholding is a:

a)

Enhancement method

b)

Segmentation technique

c)

Compression technique

d)

Feature descriptor

74.

Global thresholding fails when:

a)

Image is noisy

b)

Illumination is uniform

75.

Region-based segmentation groups pixels based on:

a)

Edges

b)

Similarity

c)

Gradients

d)

Noise

76.

Harris corner detector is based on:

a)

Gradient magnitude

b)

Eigenvalues of matrix

c)

Histogram

d)

Frequency response

77.

SIFT features are invariant to:

a)

Rotation and scale

b)

Noise only

c)

Translation only

d)

Illumination only

78.

SURF improves over SIFT by:

a)

Accuracy

b)

Speed

c)

Robustness

d)

Compression

79.

Corner points have:

a)

High intensity

b)

High variation in all directions

c)

Uniform gradient

d)

Zero gradient

80.

Feature descriptors are used for:

a)

Noise removal

b)

Image matching

c)

Image compression

d)

Image acquisition

81.

PCA reduces:

a)

Image noise

b)

Data dimensionality

c)

Image resolution

d)

Image brightness

82.

Principal components are selected based on:

a)

Mean

b)

Variance

c)

Correlation

d)

Entropy

83.

Higher eigenvalues represent:

a)

Noise

b)

Less information

c)

More information

d)

Background

84.

DCT concentrates energy in:

a)

High-frequency coefficients

b)

Low-frequency coefficients

c)

Random coefficients

d)

Edge coefficients

85.

Compression ratio is defined as:

a)

Output / Input

b)

Input / Output

c)

Bits per pixel

d)

Pixel count

86.

Medical image enhancement is critical for:

a)

Entertainment

b)

Diagnosis

c)

Compression

d)

Transmission

87.

Computer vision differs from image processing because it focuses on:

a)

Pixels

b)

Visual understanding

c)

Filtering

d)

Enhancement

88.

Image segmentation is performed before:

a)

Image acquisition

b)

Feature extraction

c)

Image sensing

d)

Sampling

89.

Sampling affects:

a)

Image resolution

b)

Image contrast

c)

Image noise

d)

Image color

90.

Quantization affects:

a)

Spatial resolution

b)

Intensity resolution

c)

Image size

d)

Image orientation

91.

Over-segmentation is common in:

a)

Edge-based methods

b)

Region growing

c)

Watershed algorithm

d)

Thresholding

92.

Morphological operations are based on:

a)

Frequency analysis

b)

Set theory

c)

Linear algebra

d)

Probability

93.

Dilation operation:

a)

Shrinks objects

b)

Expands objects

c)

Detects edges

d)

Removes noise

94.

Erosion operation:

a)

Thickens objects

b)

Smooths edges

c)

Shrinks objects

d)

Detects corners

95.

Opening operation is erosion followed by:

a)

Thresholding

b)

Dilation

c)

Closing

d)

Filtering

96.

First stage of vision pipeline:

a)

Recognition

b)

Feature extraction

c)

Image acquisition

d)

Classification

97.

Object recognition requires:

a)

Segmentation only

b)

Feature extraction and classification

c)

Filtering only

d)

Thresholding only

98.

CNNs are effective because they:

a)

Use handcrafted features

b)

Learn hierarchical features

c)

Use PCA

d)

Use DCT

99.

Pooling layers reduce:

a)

Noise

b)

Feature map size

c)

Color depth

d)

Kernel size

100.

Computer vision is widely used in:

a)

Autonomous vehicles

b)

Medical imaging

c)

Surveillance

d)

All of the above