GAMBELLA UNIVERSITY

FACULTY OF NATURAL AND COMPUTIONAL SCIENCE

DEPARTMENT OF COMPUTER SCIENCE
INDIVISUAL ASSIGNMENT:COMPUTER VISION AND IMAGE
PROCESSING
Name id
1. BIRHAN FENTIE……..……………..7774

Submission date:20 /03/2017

Submitted To Mr. Michael
GAMBELLA, ETHIOPIA
 1.write mathematical tools used in digital image processing

 Digital image processing relies heavily on mathematical tools to manipulate and analyze
image data. Here are some of the most commonly used mathematical tools:

1. Linear Algebra:

• Matrices and Vectors: Images are represented as matrices, with each element representing a
pixel's intensity value. Operations like image transformations, filtering, and color space
conversions are performed using matrix operations.

• Eigenvalues and Eigenvectors: Used in image compression techniques (like Principal

Component Analysis), dimensionality reduction, and feature extraction.

2. Calculus:

• Derivatives: Used to detect edges and other features in images, particularly in edge detection
algorithms (like the Sobel operator). Gradients are a key example.

• Integral Calculus: Applied in image smoothing, noise reduction (e.g., Gaussian blurring), and
morphological operations (like area calculations).

3. Probability and Statistics:

• Histograms: Used to analyze the distribution of pixel intensity values, aiding in contrast
enhancement, noise reduction, and segmentation.

• Statistical Models: Used to model image characteristics like texture, shape, and color for
segmentation, classification, and object recognition (e.g., using Bayes' theorem for classification).

• Bayesian Inference: Used in image restoration, noise reduction, and object detection problems.

4. Discrete Mathematics:

• Discrete Fourier Transform (DFT): Transforms an image into its frequency domain
representation, enabling efficient filtering (high-pass, low-pass), compression, and noise
reduction.

1
• Discrete Cosine Transform (DCT): A related transform used in image compression standards
like JPEG.

• Discrete Wavelet Transform (DWT): A multiresolution transform providing better spatial and
frequency localization than DFT, used in image compression, denoising, and feature extraction.

5. Set Theory and Logic:

• Morphological Operations: Set-theoretic operations (like erosion, dilation, opening, closing)

used for image analysis, modifying shapes and removing noise.

• Logical Operators: Used in image segmentation and classification, where pixels are classified
based on logical conditions.

6. Geometry and Topology:

• Geometric Transformations: Used to manipulate image geometry, including rotation, scaling,

translation, and perspective transformations (affine transformations, projective transformations).

• Shape Analysis: Analyzing shapes of objects, including boundary detection, feature extraction
(e.g., moments), and object recognition.

7. Number Theory:

• Cryptography: Applied in digital image security, including encryption and watermarking

techniques.

8. Other Mathematical Tools:

• Optimization Algorithms: Used in tasks like image registration, segmentation (e.g., level set
methods), and noise reduction.

• Numerical Methods: Applied to solve complex equations arising from image processing
operations.

 These mathematical tools are fundamental to various digital image processing techniques,
enabling analysis, enhancement, and manipulation in diverse applications. The specific tools
used depend heavily on the particular task (e.g., image compression vs. object recognition).

2
2.discrete one dimension Fourier transformation and its inverse

 The Discrete Fourier Transform (DFT) and its inverse (IDFT) are fundamental tools in
digital signal processing. They allow you to convert a signal from the time domain (where
the signal is represented as a function of time) to the frequency domain (where the signal is
represented as a function of frequency) and back again. Here's a look at the 1D DFT and
IDFT:

1. Discrete Fourier Transform (DFT)

 The DFT takes a sequence of N complex numbers, x[n] (where n = 0, 1, ..., N-1), and
transforms it into another sequence of N complex numbers, X[k] (where k = 0, 1, ..., N-1).
Each X[k] represents the amplitude and phase of a specific frequency component in the
signal.

The formula for the DFT is:

X[k] = Σ (n=0 to N-1) x[n] * exp(-j * 2π * k * n / N)

X[k] = Σ (n=0 to N-1) x[n] exp(-j 2π k n / N)

Where:

• X[k] is the k-th frequency component.

• x[n] is the n-th sample of the input signal.

• j is the imaginary unit (√-1).

• N is the number of samples in the input signal.

In simpler terms: The DFT takes each sample of your signal, multiplies it by a complex
exponential (which represents a specific frequency), and sums up the results for each frequency.

2. Inverse Discrete Fourier Transform (IDFT)

 The IDFT performs the reverse operation, converting a sequence of N complex numbers in
the frequency domain, X[k], back into the time domain sequence, x[n].

3
The formula for the IDFT is:

x[n] = (1/N) * Σ (k=0 to N-1) X[k] * exp(j * 2π * k * n / N)`

x[n] = (1/N) Σ (k=0 to N-1) X[k] exp(j 2π k * n / N)

Where:

• x[n] is the n-th sample of the reconstructed time-domain signal.

• X[k] is the k-th frequency component.

• j is the imaginary unit (√-1).

• N is the number of samples.

In simpler terms: The IDFT takes each frequency component, multiplies it by a complex
exponential (again representing a frequency), sums up the results for each time point, and then
scales the result by 1/N.

Relationship between DFT and IDFT: They are essentially inverse operations. Applying the DFT
and then the IDFT (or vice versa) should ideally reconstruct the original signal (though there can
be minor numerical errors in practice)

Practical Considerations:

• Computational Cost: Directly computing the DFT using the formula above is computationally
expensive (O(N²)). Faster algorithms like the Fast Fourier Transform (FFT) are typically used in
practice to reduce the computation time to O(N log N).

• Complex Numbers: Both the DFT and IDFT deal with complex numbers, meaning the results
have both real and imaginary parts. The magnitude of the complex number represents the
amplitude of the frequency component, and the angle (phase) represents the phase shift

 The DFT and IDFT are cornerstones of digital signal processing, used extensively in
applications such as audio processing, image processing, and communications. The FFT, a
highly efficient algorithm for computing the DFT, has made it feasible to apply these
transforms in real-time applications.

4
 3. frequency information of image

 The frequency information of an image refers to the representation of an image in the

frequency domain, rather than the spatial domain (the usual pixel representation). It
describes the rate of change of intensity levels in an image. High frequencies correspond to
sharp changes (edges, details), while low frequencies correspond to gradual changes (smooth
areas).

Here's a breakdown of how we obtain and interpret frequency information in images:

1. Obtaining Frequency Information:

 The primary method for obtaining the frequency information of an image is through the use
of the Discrete Fourier Transform (DFT) or its faster counterpart, the Fast Fourier Transform
(FFT).

• DFT/FFT: This mathematical operation transforms the image from the spatial domain (pixels)
into the frequency domain. The result is a complex-valued matrix called the frequency spectrum
or Fourier transform.

• Frequency Spectrum: The frequency spectrum doesn't represent the image in a visually
recognizable way. Instead:

 Magnitude: The magnitude of each element in the frequency spectrum represents the
strength of that particular frequency in the image. Higher magnitudes mean stronger
frequencies.

 Phase: The phase of each element represents the phase shift of that frequency. Phase
information is important for image reconstruction.

5
 2. Interpreting Frequency Information:

 • Low Frequencies (near the center of the spectrum): Represent slow changes in intensity,
corresponding to smooth areas, large regions of uniform color or intensity, and the overall
average brightness.

 • High Frequencies (further from the center of the spectrum): Represent rapid changes in
intensity, corresponding to sharp edges, fine details, textures, and noise.

 • Spatial and Frequency Relationship: A strong correlation exists between spatial and
frequency information.

 A smooth region in the image will have low frequency components, while a sharp transition
(an edge) generates high frequency components.

3. Applications of Frequency Information:

 Frequency domain analysis of images is crucial in numerous applications, including:

 Image Enhancement: Filtering specific frequency components enables tasks such as:

 Noise Reduction: Attenuating high-frequency components (noise) while preserving low-

frequency components (main image features).

 Sharpening: Enhancing high-frequency components to improve the details and edges.

 Blurring: Attenuating high-frequency components to smooth out the image and reduce noise
or details.

 Image Compression: Transform-based image compression techniques like JPEG use the
DCT (Discrete Cosine Transform), which is closely related to the DFT, to represent images
more efficiently in the frequency domain. High-frequency components, which usually have
less visual impact, are discarded or compressed more heavily.

 Feature Extraction: Frequency components can be used as features for object recognition
and image classification. Textures are often characterized by their frequency content.

6
 Image Analysis: Analyzing the frequency spectrum can provide insights into the overall
characteristics of an image, such as the presence of periodic patterns, dominant frequencies,
and the level of noise.

 In summary: The frequency information of an image, obtained using the DFT/FFT, provides
a powerful way to analyze and manipulate image content by representing it in terms of its
frequency components. Understanding this frequency representation is essential for many
advanced image processing techniques.

7
 4. image enhancement in frequency domain

 Image enhancement in the frequency domain leverages the fact that different frequencies in
an image correspond to different features. Low frequencies represent smooth areas and
overall brightness, while high frequencies represent sharp changes, edges, and fine details.
By modifying the frequency components, we can enhance various aspects of the image. This
is often done using the Discrete Fourier Transform (DFT) or its faster counterpart, the Fast
Fourier Transform (FFT).

1. Transforming to the Frequency Domain:

 The first step is to convert the image from the spatial domain (pixels) to the frequency
domain using the DFT or FFT. This results in a frequency spectrum, typically represented as
a complex-valued matrix. The magnitude of each component in this spectrum corresponds to
the amplitude of the corresponding frequency.

2. Filtering in the Frequency Domain:

 The core of frequency-domain enhancement lies in applying filters to modify the frequency
components. These filters are typically designed to attenuate (reduce) or amplify (boost)
specific frequency ranges.

• Low-pass filters: Attenuate high frequencies, resulting in image smoothing and noise reduction.
This removes high-frequency details and sharp transitions, leaving a blurred image. Examples:
Gaussian low-pass filter, Butter-worth low-pass filter.

• High-pass filters: Attenuate low frequencies, resulting in edge sharpening and detail
enhancement. This removes low-frequency information, making edges more prominent and fine
details more visible. Examples: Gaussian high-pass filter, Laplacian filter (a derivative-based
high-pass filter).

• Band-pass filters: Amplify frequencies in a specific range, allowing for the enhancement of
textures or specific features of interest.

8
• Band-reject (notch) filters: Attenuate frequencies within a specific range, useful for removing
periodic noise or artifacts.

3. Transforming Back to the Spatial Domain:

 After filtering, the inverse DFT or FFT is applied to convert the modified frequency
spectrum back into the spatial domain, producing the enhanced image.

 Specific Enhancement Techniques:

• Homomorphic filtering: Uses a logarithmic transformation to separate the illumination and

reflectance components of an image before filtering. This is often effective in improving contrast
in images with uneven lighting.

• Wiener filtering: A statistical approach to image restoration that aims to minimize the mean
square error between the original and enhanced image. It's particularly useful in removing
additive noise.

• Adaptive filtering: The filter's characteristics (e.g., cutoff frequency) vary across the image,
allowing for localized enhancement based on the image content.

Advantages of Frequency-Domain Enhancement:

• Efficiency: Certain filtering operations, such as Gaussian blurring, are computationally more
efficient in the frequency domain.

• Flexibility: Allows for precise control over the enhancement process by selectively modifying
different frequency bands.

• Noise Reduction: Frequency domain filtering is particularly effective in reducing noise,

especially periodic noise patterns

9
 Disadvantages:

• Computational Complexity (without FFT): Direct DFT computation can be slow.

• Interpretation: The frequency spectrum is not an intuitive representation of the image, making
filter design somewhat challenging.

• Phase Information: Care must be taken to preserve or appropriately handle the phase
information during filtering, as it's crucial for accurate image reconstruction.

 Frequency-domain enhancement is a powerful technique with broad applications in image

processing. The choice of filter and parameters depends significantly on the specific
characteristics of the image and the desired enhancement goal

10