Digital Image Processing Fundamentals

27th Jan 2025
13 min read
Tags:
image-processing,
Review

Updated on 28th Jan 2025
See changes

Overview

Glossary

Bit Depth: Number of bits used to represent each pixel (e.g., 8-bit = 256 shades of gray).
Contrast: Difference in intensity between light and dark areas of an image.
Intensity: Brightness of a pixel in an image.
Hue: Color attribute that distinguishes one color from another.
Saturation: Color attribute that describes intensity of color.
Luminance: Brightness of an image.
Chrominance: Color information of an image.

Color Models

Binary: Single bit representing black or white.
Grayscale: Single channel representing intensity.
RGB: (Red, Green, Blue) Additive color model
CMYK: (Cyan, Magenta, Yellow, Black) Subtractive color model
HSV: (Hue, Saturation, Value) Color model that separates intensity from color information.
YCbCr: Color model used in video compression. Y represents luminance, Cb and Cr represent chrominance.

Image Enhancement

Improves the visual quality of an image.

Spatial Domain Methods

Histogram Equalization: Improves contrast by redistributing intensity values.
Smoothing Filters: Reduces noise (e.g., mean filter, median filter).
Sharpening Filters: Highlight edges (e.g., Laplacian, Sobel, Prewitt).

Frequency Domain Methods

Fourier Transform: Converts image from spatial to frequency domain.
Low-Pass Filter (LPF): Removes high-frequency noise.
High-Pass Filter (HPF): Enhances edges and fine details.

Image Restoration

Removes noise or degradation from an image.

Noise Models

Additive Noise: Noise added to image (e.g., Gaussian, salt-and-pepper).
- Gaussian Noise: Random variation in intensity values.
- Salt-and-Pepper Noise: Randomly scattered black and white pixels.
Multiplicative Noise: Noise multiplies image (e.g., speckle).

Restoration Filters

Mean Filter: Replaces pixel with average of surrounding pixels.
Median Filter: Replaces pixel with median of surrounding pixels. Better than mean filter for salt-and-pepper noise removal (preserves edges).
Wiener Filter: Removes noise by estimating signal and noise power. Adapts to local image variations.
Inverse Filter: Restores image by deconvolving blurred image with point spread function (PSF).

Image Segmentation

Divides image into meaningful regions.

Thresholding

Converts grayscale image to binary image.

Global Thresholding: Single threshold for entire image. i.e., pixel value above threshold is considered object, below threshold is background.
Adaptive Thresholding: Threshold varies based on local image properties.

Region-Based Segmentation

Region Growing: Merges pixels with similar properties.
Split and Merge: Divides image into smaller regions, then merges regions with similar properties.

Edge Detection

Sobel Operator: Detects edges by calculating gradient magnitude.
Canny Edge Detector: Detects edges by finding local maxima of gradient magnitude. Steps:
1. Apply Gaussian filter to smooth image. (Smoothing)
2. Calculate gradient magnitude and direction.
3. Apply non-maximum suppression to thin edges.
4. Apply hysteresis thresholding to detect strong and weak edges. (Edge tracking)
Prewitt Operator: Detects edges by approximating gradient magnitude.
Laplacian Operator: Detects edges by finding zero crossings.

Morphological Operations

Modifies image based on shape of objects. Shape-based analysis.

Erosion: Shrinks objects in binary image. Removes pixels on object boundaries.
Dilation: Expands objects in binary image. Adds pixels to object boundaries.
Opening: Erosion followed by dilation. Removes small objects.
Closing: Dilation followed by erosion. Fills small holes.

Image Compression

Reduces size of image data for storage or transmission.

Lossless Compression

Retains original quality (i.e., PNG, TIFF).

Run-Length Encoding (RLE): Replaces consecutive pixels with count and value.
Huffman Coding: Variable-length encoding based on frequency of symbols.

Lossy Compression

Sacrifices quality for smaller file size (i.e., JPEG).

Discrete Cosine Transform (DCT): Converts image (spatial data) to frequency domain.
Quantization: Reduces precision of DCT coefficients.
Entropy Encoding: Variable-length encoding based on probability of symbols.

Quiz

Question 0x01

Approaches to image processing that work directly on the pixels of incoming image work in ____________.

Spatial domain
Inverse transformation
Transform domain
None of the Mentioned

Answer 0x01

Spatial domain. The spatial domain refers to the image plane itself, where the image is defined by its pixel values.

Question 0x02

Which of the following is the primary objective of sharpening of an image?

Decrease the brightness of the image
Increase the brightness of the image
Highlight fine details in the image
Blurring the image

Answer 0x02

Sharpening an image aids in highlighting small features in the image or enhancing details that have become blurred owing to factors such as noise addition.

Question 0x03

Which of the following makes an image difficult to enhance?

Dynamic range of intensity levels
High noise
Narrow range of intensity levels
All the mentioned

Answer 0x03

Dynamic range of intensity levels, High noise and Narrow range of intensity levels make it difficult to enhance an image.

Question 0x04

Which of the following is the application of Histogram Equalization?

Blurring
Contrast adjustment
Image enhancement
None of the Mentioned

Answer 0x04

Dark images are usually Enhancement using Image enhancement.

Question 0x05

What is/are the resultant image of a smoothing filter?

Image with reduced sharp transitions in gray levels
Image with high sharp transitions in gray levels
None of the mentioned
All the mentioned

Answer 0x05

Explanation: Smoothing filters reduce noise in random noise, which features sharp gray level transitions.

Question 0x06

___________ is/are the feature(s) of a highpass filtered image.

An overall sharper image
Have less gray-level variation in smooth areas
Emphasized transitional gray-level details
All of the mentioned

Answer 0x06

Explanation: A highpass filter reduces the low frequency to reduce gray-level variance in smooth sections while allowing high frequencies to emphasize transitional gray-level details for a clearer image.

Question 0x07

Which of the following techniques is used for edge detection in images?

Histogram equalization
Fourier transform
Sobel operator
Dilation

Answer 0x07

Sobel operator. Histogram equalization is used for contrast enhancement, Fourier transform is used for frequency domain analysis, and Dilation is used for morphological operations.

Question 0x08

What is the purpose of the Fourier transform in image processing?

To detect edges in an image
To convert an image from the spatial domain to the frequency domain
To compress an image
To apply a filter to an image

Answer 0x08

To convert an image from the spatial domain to the frequency domain.

Question 0x09

Which of the following is the correct definition of “morphological operations” in image processing?

Operations based on the pixel values
Operations based on the shape or structure of objects in the image
Operations based on color histograms
Operations based on frequency domain transformations

Answer 0x09

Operations based on the shape or structure of objects in the image. Morphological operations are used to modify the shape of objects in an image based on their structure.

Question 0x0a

In the context of image segmentation, what is thresholding used for?

To segment an image based on pixel intensity
To detect edges
To compress the image
To apply filters to the image

Answer 0x0a

To segment an image based on pixel intensity. Thresholding is used to convert a grayscale image into a binary image by segmenting the image based on pixel intensity. Sobel, Canny, and Prewitt operators are used for edge detection. Filters are used for image enhancement and restoration.

Question 0x0b

Which of the following techniques can be used for image resizing?

Bilinear interpolation
Sobel edge detection
Fast Fourier Transform
Median filtering

Answer 0x0b

Bilinear interpolation. Bilinear interpolation is a method used for image resizing that calculates new pixel values based on the weighted average of surrounding pixels. Sobel edge detection is used for edge detection, Fast Fourier Transform is used for frequency domain analysis, and Median filtering is used for noise reduction.

Question 0x0c

Which of the following algorithms is commonly used for image compression?

Canny edge detector
Run-length encoding
K-means clustering
Gradient descent

Answer 0x0c

Run-length encoding. Run-length encoding is a lossless compression algorithm commonly used for image compression. The Canny edge detector is used for edge detection, K-means clustering is used for image segmentation, and Gradient descent is an optimization algorithm.

Question 0x0d

Which method is commonly used for removing salt-and-pepper noise from an image?

Gaussian smoothing
Median filtering
Histogram equalization
Edge detection

Answer 0x0d

Median filtering. Median filtering is commonly used for removing salt-and-pepper noise from an image by replacing pixel values with the median of neighboring pixels. Gaussian smoothing is used for noise reduction and blurring, Histogram equalization is used for contrast enhancement, and Edge detection is used for detecting edges in an image.

Question 0x0e

What does the term “image thresholding” refer to?

Converting a color image to grayscale
Detecting edges in an image
Converting an image into a binary image based on intensity levels
Enhancing image contrast

Answer 0x0e

Converting an image into a binary image based on intensity levels. Image thresholding is the process of converting a grayscale image into a binary image by segmenting the image based on intensity levels. Converting a color image to grayscale is known as color space conversion.

Question 0x0f

Which of the following techniques is used to obtain a sharper image by highlighting high-frequency components?

Low-pass filtering
High-pass filtering
Median filtering
Histogram equalization

Answer 0x0f

High-pass filtering. High-pass filtering is used to highlight high-frequency components in an image, resulting in a sharper image. Low-pass filtering is used to remove high-frequency noise, Median filtering is used for noise reduction, and Histogram equalization is used for contrast enhancement.

Question 0x10

In which of the following situations is a bilateral filter most useful?

Reducing Gaussian noise
Blurring an image
Preserving edges while smoothing an image
Detecting edges

Answer 0x10

Preserving edges while smoothing an image. A bilateral filter is most useful when you want to smooth an image while preserving edges. It achieves this by considering both spatial and intensity differences between pixels. Reducing Gaussian noise is typically done with a Gaussian filter, Even though bilateral filter uses Gaussian kernel, it is not used for reducing Gaussian noise.

Question 0x11

What is a common application of the Fast Fourier Transform (FFT) in image processing?

Image compression
Edge detection
Frequency domain filtering
Morphological transformation

Answer 0x11

Frequency domain filtering. The FFT is commonly used in image processing to convert an image from the spatial domain to the frequency domain, where frequency domain filtering can be applied to enhance or suppress specific frequencies in the image. Image compression, Edge detection, and Morphological transformation are not direct applications of the FFT.

Question 0x12

Which type of noise is best reduced using an adaptive filter?

Gaussian noise
Salt-and-pepper noise
Speckle noise
Periodic noise

Answer 0x12

Gaussian noise. Adaptive filters are best suited for reducing Gaussian noise, which is random and can vary in intensity. Median filters are more effective for reducing salt-and-pepper noise, and Wiener filters are commonly used for reducing speckle noise. Periodic noise can be reduced using frequency-domain filtering techniques.

Question 0x13

Which of the following is NOT an image restoration technique?

Inverse filtering
Wiener filtering
Image thresholding
Blind deconvolution

Answer 0x13

Image thresholding. Image thresholding is a segmentation technique used to convert a grayscale image into a binary image based on intensity levels. Inverse filtering, Wiener filtering, and Blind deconvolution are all image restoration techniques used to recover the original image from a degraded or noisy version.

Question 0x14

Which of the following transforms is most commonly used for edge detection?

Hough Transform
Discrete Cosine Transform
Wavelet Transform
Laplacian Transform

Answer 0x14

Laplacian Transform. The Laplacian transform is commonly used for edge detection in image processing. The Hough Transform is used for line detection, the Discrete Cosine Transform is used for image compression, and the Wavelet Transform is used for multi-resolution analysis.

Question 0x15

What is the purpose of a Gaussian filter in image smoothing?

To enhance contrast
To preserve edges while reducing noise
To remove high-frequency components
To segment objects in an image

Answer 0x15

To preserve edges while reducing noise. A Gaussian filter is used in image smoothing to reduce noise while preserving edges. It achieves this by convolving the image with a Gaussian kernel that reduces high-frequency components while maintaining the overall structure of the image.

Question 0x16

Which of the following is used for noise removal while preserving image details?

High-pass filter
Adaptive filter
Histogram equalization
Erosion

Answer 0x16

Adaptive filter. Adaptive filters are used for noise removal while preserving image details by adjusting the filter coefficients based on the local image characteristics. High-pass filters are used to enhance edges and fine details, Histogram equalization is used for contrast enhancement, and Erosion is a morphological operation used to shrink objects in binary images.

Question 0x17

Which of the following methods is most suitable for detecting circular objects in an image?

Sobel edge detection
Hough Transform
Fourier Transform
Laplacian filtering

Answer 0x17

Hough Transform. The Hough Transform is most suitable for detecting circular objects in an image by representing them as points in parameter space. Sobel edge detection is used for edge detection, Fourier Transform is used for frequency domain analysis, and Laplacian filtering is used for edge detection.

Question 0x18

Which of the following is a common application of morphological operations in image processing?

Edge detection
Noise reduction
Shape analysis
Color correction

Answer 0x18

Shape analysis. Morphological operations are commonly used for shape analysis in image processing, where they modify the shape of objects based on their structure. Edge detection is performed using edge detection algorithms, Noise reduction is achieved using filters, and Color correction is done using color space transformations.