Back to Subject

Unit 1: Fundamentals of Image Processing, Image Enhancement, and Noise removal and filtering - Practice Quiz

INT345 — Computer Vision 60 Questions
0 Correct 0 Wrong 60 Left
0/60

1 What is a digital image essentially composed of?

Introduction to image processing Easy
A. A continuous analog signal
B. A 2D array of discrete picture elements called pixels
C. A series of text strings and metadata
D. A continuous one-dimensional audio waveform

2 What is the primary goal of digital image processing?

Introduction to image processing Easy
A. To reduce the physical size of a computer monitor
B. To improve image data for human interpretation or machine perception
C. To generate 3D models from scratch without any input data
D. To convert 2D images directly into audio files

3 In terms of inputs and outputs, how does Image Processing generally differ from Computer Vision?

key differences between image processing and computer vision Easy
A. There is no difference; the terms mean exactly the same thing.
B. Image processing only deals with 3D models, while computer vision deals with 2D images.
C. Image processing outputs text, while computer vision outputs images.
D. Image processing outputs an image, while computer vision outputs understanding or decisions.

4 Which of the following tasks is considered a Computer Vision task rather than basic Image Processing?

key differences between image processing and computer vision Easy
A. Converting a color image to grayscale
B. Object detection and facial recognition
C. Adjusting image brightness and contrast
D. Sharpening a blurry photograph

5 Which of the following is a common medical application of image processing?

applications of image processing and computer vision Easy
A. Compiling software code for hospital databases
B. Enhancing X-ray, CT, and MRI scans for better diagnosis
C. Sending text messages between doctors
D. Creating financial spreadsheets for hospital administration

6 Automated face recognition used to unlock modern smartphones is an application of which field?

applications of image processing and computer vision Easy
A. Natural Language Processing
B. Audio Signal Processing
C. Computer Vision
D. Database Management

7 Which standard image file format is widely known for utilizing lossy compression to reduce file size?

image file formats Easy
A. TIFF
B. PNG
C. BMP
D. JPEG

8 What does the acronym PNG stand for in the context of image file formats?

image file formats Easy
A. Public Network Grid
B. Picture Network Group
C. Pixel Node Graphics
D. Portable Network Graphics

9 What does 'contrast' refer to in a digital image?

contrast enhancement Easy
A. The specific file extension (e.g., .jpg vs .png) of the image
B. The physical size of the printed image
C. The difference in luminance or color that makes an object distinguishable from its background
D. The amount of disk space the image occupies

10 Which technique is primarily used to make a dark, low-visibility image appear clearer by expanding its range of pixel intensities?

contrast enhancement Easy
A. Contrast enhancement
B. Image compression
C. Image cropping
D. Format conversion

11 What does an image histogram graphically represent?

histogram equalization and specification Easy
A. The size of the image file in kilobytes
B. The number of unique objects detected within the image
C. The physical dimensions of the image in pixels
D. The frequency distribution of pixel intensity values across the image

12 What is the main objective of histogram equalization?

histogram equalization and specification Easy
A. To add artificial colors to a grayscale image
B. To achieve a more uniform distribution of pixel intensities to enhance overall contrast
C. To change the image file format from JPEG to PNG
D. To reduce the image resolution and save storage space

13 What characterizes 'Salt and Pepper' noise in a digital image?

image noise and its types Easy
A. A digital watermark superimposed on the image
B. Random occurrences of pure black and pure white pixels
C. A type of blur caused by camera motion
D. Color distortion caused by poor lighting conditions

14 Which type of image noise is often modeled mathematically using a bell-shaped (normal) distribution?

image noise and its types Easy
A. Gaussian noise
B. Impulse noise
C. Speckle noise
D. Quantization noise

15 What typically causes noise in digital photographs taken in low light?

image noise and its types Easy
A. Using a high-resolution 4K monitor
B. Saving the image as a lossless PNG file
C. High sensor sensitivity (ISO) and electronic interference
D. Taking a photograph of a perfectly smooth surface

16 What does a spatial domain filter operate directly on?

spatial domain filtering techniques Easy
A. The geometric coordinates of detected objects
B. The color metadata of the image file
C. The individual pixels of the image
D. The Fourier transform coefficients of the image

17 Which spatial filter is highly effective at removing salt-and-pepper noise while preserving sharp edges?

spatial domain filtering techniques Easy
A. Median filter
B. Mean (Average) filter
C. High-pass filter
D. Laplacian filter

18 A spatial smoothing (low-pass) filter is primarily used for what purpose?

spatial domain filtering techniques Easy
A. Detecting sharp edges in the image
B. Increasing the global image contrast
C. Blurring the image and reducing noise
D. Compressing the image file size

19 Which mathematical transform is used to convert an image from the spatial domain into the frequency domain?

frequency domain filtering Easy
A. Distance Transform
B. Affine Transform
C. Fourier Transform
D. Hough Transform

20 In the frequency domain, what do high frequencies generally correspond to in an image?

frequency domain filtering Easy
A. Smooth backgrounds and uniform color areas
B. The physical dimensions (width and height) of the image
C. Edges, fine details, and sharp transitions in intensity
D. The overall average brightness of the image

21 Which of the following scenarios best illustrates the transition from Image Processing to Computer Vision in an automated pipeline?

key differences between image processing and computer vision Medium
A. Applying a median filter to remove noise, followed by using a Sobel operator to detect edges.
B. Converting an RGB image to grayscale, and then applying histogram equalization to improve contrast.
C. Compressing a high-resolution image into a JPEG format, followed by transmitting it over a network.
D. Enhancing the contrast of a medical X-ray, followed by an algorithm that identifies and classifies a tumor.

22 In a digital camera's software, which sequence correctly identifies the roles of Image Processing (IP) and Computer Vision (CV)?

key differences between image processing and computer vision Medium
A. IP: Recognizing a face for autofocus; CV: Applying white balance.
B. IP: Estimating the depth of the scene; CV: Adjusting the brightness.
C. IP: Applying white balance and noise reduction; CV: Detecting faces to draw bounding boxes.
D. IP: Compressing the image to JPEG; CV: Sharpening the image.

23 If a continuous image is digitized such that the spatial resolution is increased while the amplitude resolution (quantization levels) is decreased, what is the most likely visual effect on the resulting digital image?

Introduction to image processing Medium
A. The image will have smoother color transitions but lose high-frequency edge details.
B. The image will become blurry and exhibit a checkerboard pattern.
C. The image will have sharper details but will exhibit false contouring.
D. The image will appear exactly the same as long as the file size remains constant.

24 In the context of Optical Character Recognition (OCR), which of the following represents a Computer Vision task rather than a preliminary Image Processing task?

applications of image processing and computer vision Medium
A. Mapping a segmented connected component to the letter 'A' using a trained model.
B. Applying a morphological closing operation to bridge gaps in broken text strokes.
C. Binarizing the scanned document using Otsu's thresholding.
D. Correcting the skew of the scanned document.

25 Which of the following applications relies primarily on Image Processing techniques without necessarily requiring Computer Vision?

applications of image processing and computer vision Medium
A. Facial recognition for smartphone unlocking.
B. Automated reading of vehicle license plates at a toll booth.
C. Restoration of an old, degraded photograph by removing scratches.
D. Autonomous driving obstacle avoidance.

26 A researcher is storing images containing sharp text and line drawings for an archival dataset. They need a format that avoids artifacts around the text edges and supports a lossless workflow. Which of the following formats is the most appropriate?

image file formats Medium
A. PNG
B. GIF
C. JPEG
D. Lossy WebP

27 Which characteristic makes the TIFF file format highly suitable for professional medical and satellite imaging?

image file formats Medium
A. It is highly compressed, resulting in the smallest file sizes for web transmission.
B. It inherently restricts the color depth to 8-bit, standardizing image analysis.
C. It is an animation-only format that allows for the viewing of 3D medical scans.
D. It supports high bit-depths, lossless compression, and can store multiple images (pages) in a single file.

28 An image captured in a low-light environment appears very dark. To enhance the image using a power-law (Gamma) transformation , what should be the value of ?

contrast enhancement Medium
A.
B.
C.
D.

29 What is the primary purpose of applying a Log transformation to an image?

contrast enhancement Medium
A. To compress the dynamic range of an image, expanding dark pixels and compressing bright pixels.
B. To threshold the image into a binary format.
C. To increase the contrast of the bright pixels while compressing the dark pixels.
D. To invert the image intensities, producing a negative image.

30 In piecewise linear transformation, what is the effect of setting the coordinates and such that and , (where is the number of intensity levels)?

contrast enhancement Medium
A. It performs intensity-level slicing.
B. It applies contrast stretching.
C. It performs a thresholding operation, creating a binary image.
D. It reverses the image, creating a negative.

31 If global histogram equalization is applied to an image whose histogram is heavily clustered at the lower intensity levels (dark image), what will be the characteristic of the resulting output image's histogram?

histogram equalization and specification Medium
A. It will be shifted perfectly to the higher intensity levels, making the image overly bright.
B. It will become completely flat with exactly equal numbers of pixels at every intensity level.
C. It will remain clustered at the lower intensity levels, but the contrast will decrease.
D. It will be stretched across the entire intensity range, though not perfectly flat due to discrete pixel values.

32 Under what circumstance is Histogram Specification (Matching) preferred over standard Histogram Equalization?

histogram equalization and specification Medium
A. When the image contains salt-and-pepper noise that needs to be removed.
B. When we need the output image to have a specific predefined histogram shape rather than a uniform one.
C. When the image requires maximum global contrast enhancement without any specific target distribution.
D. When processing time is strictly limited, as specification is computationally faster than equalization.

33 Let be the probability density function of the input image and be the transformation function for histogram equalization. What mathematical property must satisfy to ensure the output image intensities retain their logical order without inversion?

histogram equalization and specification Medium
A. must be a periodic function.
B. must be a step function.
C. must be a strictly decreasing function.
D. must be a monotonically increasing function.

34 Which type of noise is characterized by a probability density function (PDF) that consists of two distinct impulses at the extremes of the intensity range?

image noise and its types Medium
A. Uniform noise
B. Salt-and-pepper (impulse) noise
C. Gaussian noise
D. Speckle noise

35 In medical ultrasound imaging, images are often degraded by a granular noise pattern. This noise is multiplicative in nature. What type of noise is this?

image noise and its types Medium
A. Speckle noise
B. Gaussian noise
C. Salt-and-pepper noise
D. Periodic noise

36 Why is a median filter generally preferred over an average (mean) filter for removing salt-and-pepper noise?

spatial domain filtering techniques Medium
A. Because the average filter increases the dynamic range of the noise.
B. Because the median filter alters all pixel values in the image uniformly.
C. Because the median filter is a linear filter and therefore computationally faster.
D. Because the average filter severely blurs the edges, whereas the median filter preserves edge sharpness while replacing the extreme noise values.

37 Which of the following spatial filters would you apply to highlight fine details and enhance blurred edges in an image?

spatial domain filtering techniques Medium
A. A median filter.
B. A matrix filled entirely with .
C. A Gaussian smoothing filter.
D. A Laplacian filter.

38 In unsharp masking, the sharpened image is obtained using the formula: . How is the generated?

spatial domain filtering techniques Medium
A. By adding a Gaussian noise pattern to the original image.
B. By applying a high-pass filter directly to the original image.
C. By taking the first derivative of the original image.
D. By subtracting a blurred (low-pass filtered) version of the image from the original image.

39 When applying an Ideal Lowpass Filter (ILPF) in the frequency domain, what visible artifact is most commonly introduced in the spatial domain output?

frequency domain filtering Medium
A. False contouring
B. Multiplicative speckle
C. Ringing effect (Gibbs phenomenon)
D. Salt-and-pepper noise

40 In the frequency domain, what structural components of an image generally map to the high-frequency regions of the Fourier spectrum?

frequency domain filtering Medium
A. Slowly varying color transitions.
B. Edges, sharp transitions, and noise.
C. Smooth gradients and uniform backgrounds.
D. The average brightness (DC component) of the image.

41 Which of the following fundamentally distinguishes the architectural endpoint of a Computer Vision (CV) system from a classical Digital Image Processing (DIP) system in the context of the semantic gap?

Key differences between image processing and computer vision Hard
A. DIP minimizes the semantic gap by strictly using lossless compression, while CV uses lossy compression.
B. CV systems map multi-dimensional arrays to 1D vectors, whereas DIP systems always maintain 3D topological outputs.
C. DIP systems utilize spatial domain operations exclusively, whereas CV systems rely entirely on frequency domain transforms.
D. DIP systems output an enhanced scalar matrix, whereas CV systems output highly abstracted semantic descriptors or decisions.

42 Given a continuous image with intensity values in the range and probability density function , histogram equalization applies the transformation . By applying the fundamental theorem of calculus and probability transformation principles, what is the resulting probability density function of the output image?

Histogram equalization and specification Hard
A. for
B. for
C.
D.

43 Why does discrete histogram equalization rarely result in a perfectly flat (uniform) histogram in practice, unlike the continuous formulation?

Histogram equalization and specification Hard
A. Because discrete transformations cannot break pixel groupings; all pixels with a specific input intensity must map to exactly the same output intensity .
B. Because rounding errors in the discrete integral cause high-frequency attenuation.
C. Because spatial filtering is required before discrete equalization can map non-linear intensities.
D. Because the number of output bins is always strictly less than the number of input bins.

44 Consider a power-law (Gamma) transformation applied to an image with intensity range . If a nested transformation is applied such that followed by , which condition must hold for the final image to strictly compress the dynamic range of the dark regions while expanding the dynamic range of the bright regions?

Contrast enhancement Hard
A.
B.
C.
D.

45 An image is degraded by noise whose probability density function is given by for , and otherwise. This noise model is particularly useful for characterizing noise in which of the following imaging modalities?

Image noise and its types Hard
A. Magnetic Resonance Imaging (MRI) and range imaging, representing Rayleigh noise.
B. Laser imaging and coherent synthetic aperture radar (SAR), representing Erlang (Gamma) noise.
C. Photon-limited X-ray imaging, representing Poisson noise.
D. Thermal imaging circuits, representing Gaussian (electronic) noise.

46 An alpha-trimmed mean filter is applied to an neighborhood. It operates by sorting the pixels, deleting the lowest and highest values, and averaging the rest. If , what does this filter mathematically reduce to?

Spatial domain filtering techniques Hard
A. A Median filter
B. An Arithmetic Mean filter
C. A Min filter
D. A Max filter

47 The Unsharp Masking technique for image sharpening is defined as , where is a blurred version of . If the weighting constant , this process is specifically known as:

Spatial domain filtering techniques Hard
A. Laplacian of Gaussian filtering
B. Sobel gradient extraction
C. Homomorphic filtering
D. High-boost filtering

48 An Ideal Lowpass Filter (ILPF) in the frequency domain is a radial cylinder function. Why does applying an ILPF severely introduce 'ringing' artifacts in the spatial domain?

Frequency domain filtering Hard
A. Because the sharp cutoff acts as a spatial domain high-pass filter, accentuating edges periodically.
B. Because removing high frequencies violates the Nyquist-Shannon sampling theorem, causing aliasing that manifests as rings.
C. Because the inverse Fourier transform of a 2D cylinder is a 2D sinc function (or Bessel function of the first kind), which has infinite oscillating side lobes.
D. Because the ILPF introduces non-linear phase shifts at the cutoff frequency.

49 Consider a Butterworth Lowpass Filter (BLPF) of order , defined as . As the order , which phenomenon becomes most prominent when filtering a natural image?

Frequency domain filtering Hard
A. The transition band of the filter becomes perfectly flat, rendering the filter equivalent to an identity transform.
B. The filter perfectly preserves both high and low frequencies, eliminating any blurring.
C. The BLPF approaches an Ideal Lowpass Filter, resulting in pronounced spatial ringing artifacts.
D. The BLPF approaches a Gaussian Lowpass Filter, ensuring zero ringing artifacts.

50 Homomorphic filtering is designed to improve the appearance of an image by simultaneous intensity range compression and contrast enhancement. It achieves this by modeling the image as the product of illumination and reflectance . What is the exact sequence of operations in homomorphic filtering?

Frequency domain filtering Hard
A. Exponential DFT Low-pass filter IDFT Logarithm
B. Logarithm DFT Low-pass filter IDFT Exponential
C. Logarithm DFT High-frequency emphasis filter IDFT Exponential
D. DFT Logarithm High-pass filter Exponential IDFT

51 In the baseline JPEG compression standard, the spatial domain image is divided into blocks and transformed using the Discrete Cosine Transform (DCT). Which step in the JPEG pipeline is primarily responsible for its lossy nature and achieves the most significant compression?

Image file formats Hard
A. Element-wise division of the DCT coefficients by a quantization matrix followed by rounding.
B. The application of the forward Discrete Cosine Transform to the spatial blocks.
C. Huffman or Arithmetic entropy coding of the zig-zag scanned AC coefficients.
D. Differential Pulse Code Modulation (DPCM) applied to the DC coefficients.

52 Salt-and-pepper noise is a bipolar impulse noise where pixels are randomly degraded to the maximum (salt) or minimum (pepper) intensity values. If the probability of salt is and pepper is , what is the theoretical spatial domain limit of a purely linear spatial filter (like a mean filter) in removing this noise at high noise densities ()?

Image noise and its types Hard
A. Linear filters fail because they distribute the extreme outlier values across the neighborhood, causing pervasive blurring without eliminating the impulses.
B. Linear filters will successfully remove the noise but cause aliasing in the frequency domain.
C. Linear filters convert bipolar impulse noise into Gaussian noise, which can then be easily removed using a secondary linear filter.
D. Linear filters are mathematically proven to perfectly eliminate impulse noise if the kernel size strictly equals .

53 The continuous Laplacian is a linear, isotropic second-derivative operator defined as . When discretizing this operator for digital image processing using a kernel that incorporates diagonal neighbors, what must the sum of all coefficients in the mask equal, and why?

Spatial domain filtering techniques Hard
A. -1, to ensure the high-frequency components are strictly inverted before edge masking.
B. Eight, because there are exactly 8 neighbors contributing to the central pixel.
C. One, to preserve the average background intensity of the image during convolution.
D. Zero, because the derivative of a constant intensity area must yield zero.

54 According to the 2D Nyquist-Shannon sampling theorem, undersampling a continuous image containing high spatial frequencies leads to aliasing. In digital images, aliasing frequently manifests as visually distinct, low-frequency artifact patterns that do not exist in the original scene. What is the technical term for these patterns?

Introduction to image processing Hard
A. Moiré patterns
B. Gibbs phenomenon
C. Mach bands
D. Speckle artifacts

55 Bit-plane slicing breaks down an 8-bit image into 8 binary planes. Plane 7 contains the Most Significant Bits (MSB) and Plane 0 contains the Least Significant Bits (LSB). If a complex natural image is subjected to bit-plane slicing, what is the typical visual characteristic of Plane 0?

Contrast enhancement Hard
A. It appears as visually unstructured, random noise because it represents the finest, lowest-magnitude intensity variations.
B. It contains the global illumination gradient, appearing as a smooth transition from dark to light.
C. It resembles a high-contrast binary silhouette of the primary objects in the image.
D. It perfectly retains the sharpest structural edges of the objects but lacks any gray-level shading.

56 In an Automated Optical Inspection (AOI) pipeline for printed circuit boards (PCBs), a system must identify microscopic hairline fractures in copper traces while ignoring uniform, large variations in solder mask illumination. Which sequence of operations represents the most robust algorithmic synthesis for this task?

Applications of image processing and computer vision Hard
A. Apply an arithmetic mean filter followed by global thresholding.
B. Apply global histogram equalization followed by an Ideal Lowpass Filter.
C. Apply a Gaussian Lowpass Filter followed by a morphological erosion with a large structuring element.
D. Apply a Top-Hat morphological transform followed by Canny edge detection.

57 To properly display or design filters for the 2D Discrete Fourier Transform (DFT) of an image , it is common practice to shift the zero-frequency (DC) component to the center of the frequency grid. Mathematically, how is this achieved strictly in the spatial domain prior to the transform?

Frequency domain filtering Hard
A. By taking the absolute value of and dividing by its maximum
B. By convolving with a Dirac delta function
C. By multiplying by
D. By multiplying by

58 The Tagged Image File Format (TIFF) relies on a specific structural hierarchy to support immense flexibility, allowing it to contain multiple images, varying color spaces, and different compression algorithms (LZW, ZIP, etc.) in one file. What is the core data structure inside a TIFF that enables this flexibility?

Image file formats Hard
A. A globally linked Huffman tree defining the color palette.
B. A standardized header block restricted to 1024 bytes containing rigid EXIF nodes.
C. Image File Directories (IFDs) containing arrays of tagged entries.
D. A strict continuous multi-channel interleaved byte stream.

59 The Sobel edge detection operator utilizes a kernel to approximate the gradient of the image. The -direction kernel uses values . What is the theoretical justification for using a weight of '2' for the center row, as opposed to a uniform weight of '1' used in the Prewitt operator?

Spatial domain filtering techniques Hard
A. It provides greater smoothing along the axis orthogonal to the derivative, reducing the effect of high-frequency noise on edge detection.
B. It is required to make the kernel separable into 1D filters, which is mathematically impossible for the Prewitt operator.
C. It ensures that the convolution sum reaches exactly zero, which Prewitt kernels fail to achieve.
D. It strictly satisfies the Nyquist criterion for discrete spatial derivatives, preventing edge aliasing.

60 A continuous image undergoes a contrast stretching transformation defined by a piecewise linear function. If the function maps the intensity range to , and it is strictly true that while and , what does this specific transformation behave as?

Contrast enhancement Hard
A. A hard binary thresholding function at threshold .
B. A purely linear negation (image inversion) function.
C. A histogram equalization function matching a uniform distribution.
D. An identity transformation leaving the image unchanged.
Next MCQ
Unit 2