Scribd
Upload
48 views12 pages

Fourier Transforms in Image Processing

The document discusses different types of image transforms including orthogonal transforms like Fourier transforms and discrete cosine transforms, and non-sinusoidal transforms like Haar transforms. It also discusses some practical applications of image transforms like converting spatial information to frequency domain information. Finally, it provides an introduction to Fourier transforms, describing them as unitary transforms that can describe a signal as a sum of sine and cosine waves of different frequencies weighted by complex functions.

Uploaded by

siddharth.20bcun006
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
48 views12 pages

Fourier Transforms in Image Processing

The document discusses different types of image transforms including orthogonal transforms like Fourier transforms and discrete cosine transforms, and non-sinusoidal transforms like Haar transforms. It also discusses some practical applications of image transforms like converting spatial information to frequency domain information. Finally, it provides an introduction to Fourier transforms, describing them as unitary transforms that can describe a signal as a sum of sine and cosine waves of different frequencies weighted by complex functions.

Uploaded by

siddharth.20bcun006
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd

Digital Image Processing

Types of image transforms


CONTENTS

1. Types of Image Transforms.


2. Some practical applications of Image Transforms
3. Basis for a Transform.
4. Introduction to Fourier Transforms.
5. Discrete Fourier Transforms.
6. Numerical
Basic steps for filtering in the frequency domain
Basic steps for filtering in the frequency domain

IMAGE IN FREQUENCY
SPATIAL COMPONENTS IN
DOMAIN FREQUENCY DOMAIN
TYPES OF Image TRANSFORMS
• Essentially there are two categories of transforms available
1. Orthogonal: uses basic functions that are sinusoidal in nature
 Fourier Transforms
 Discrete Cosine Transforms
2. Non-sinusoidal: uses basic functions that are non-sinusoidal in nature
 Haar Transforms
 Walsh Transforms
 Hadamard Transforms

Note: Fourier transform methods are useful in converting a spatial description of an


image into another description of the information in terms of image frequency
components.
Some practical applications OF Image TRANSFORMS

• They convert spatial information into frequency domain information.

• Useful of getting information about sampling.

• Help to design faster algorithms.

• Spatial frequency transforms helps in assessing image quality.


Basis for a transform
• Let f(x) be the vector of the size N, with input samples f=[f(0),f(1)…F(N-1)]^T
Where T is the transformation applied to this input sequence to yield the resultant
vector of the form g(x).
• If so g(x) is also a vector of the size N-1.
• This transformation vector can be represented as:

• Here g(u) is the transformation of f(x) and matrix T(u,x) is called the forward
transformation kernel .
• Similarly the inverse transformation can be applied to obtain the original
vector. The inverse transformation is of the form:
Basis for a transform
• As image are 2D signals, 2D transforms are required.
• For an N X N image f(x,y), the forward and inverse transforms are as follows:

• Here T(u,v,x,y) and I (x,y.u,v) are called forward and inverse transformation
kernels respectively.
• In any image transforms the main task is to find suitable kernels. Once it is
designed the transformation can be carried out on the image.
Introduction to Fourier transform
• Fourier transform is a unitary transform and is based on the concept of basis
vectors and basis images.

• a unitary transformation is a transformation that preserves the inner product:


the inner product of two vectors before the transformation is equal to their
inner product after the transformation.

• Important feature of Fourier transform is that irrespective of the type of the


signal, it is possible to describe the signal as the sum of a collection of sine
and/or cosine waves of different frequencies and multiplied by the weighted
functions.
Introduction to Fourier transform

• Where j = √ -1
• Fourier transform treats all the image data as complex number
Introduction to Fourier transform
• Similarly for given F(u), inverse transformation can be applied to get f(x)
• Mathematically inverse Fourier transform can be expressed as:
Introduction to Fourier transform

Common questions

Powered by AI

The inverse Fourier transform is applied to retrieve an image from its frequency components by mathematically combining the sine and cosine waves of varied frequencies back into the original spatial domain representation. This involves applying the inverse transformation kernel to the frequency domain data to reconstruct the spatial data .

The Fourier transform enables the conversion of spatial descriptions of images into frequency components by describing any signal as a sum of sine and cosine waves of different frequencies, each multiplied by a weighted function. This transform translates the spatial variation of pixel values in an image into frequency data, highlighting periodic structures within the image .

Basic steps for filtering an image in the frequency domain include transforming the image from the spatial domain to the frequency domain using an image transform like the Fourier transform. Once in the frequency domain, filtering operations are applied to manipulate the frequency components accordingly, and finally, the transformed image is converted back to the spatial domain using an inverse transform to obtain the filtered image .

Unitary transformations, such as the Fourier transform, are significant because they preserve the inner product, ensuring that the inner product of two vectors remains the same before and after the transformation. This property is crucial for maintaining the original energy and relationships within the data, facilitating accurate reconstruction of the original signal from its frequency components .

Transformation kernels are matrices used in both forward and inverse image transforms. In forward transformation, the kernel is applied to the input vector or image to convert it into a frequency domain representation. Conversely, the inverse transformation kernel is applied to convert the frequency domain representation back to its original spatial form. Therefore, choosing appropriate transformation kernels is crucial for accurate and efficient image processing .

Non-sinusoidal transforms like Haar Transforms are useful for certain applications due to their mathematical simplicity and the ability to represent data with abrupt changes and compact representations. They often require less computation compared to sinusoidal transforms, making them suitable for applications involving data compression and noise reduction in images .

Image transforms have several practical applications, such as converting spatial information into frequency domain information, which helps in analyzing the image data in terms of frequency components. They are useful in sampling analysis, designing faster algorithms, and assessing image quality by analyzing spatial frequency components .

Fourier transforms handle the complex nature of image data by treating all image data as complex numbers. This allows them to effectively represent both the magnitude and phase of the image frequency components, which is essential for accurate image reconstruction and manipulation in the frequency domain .

The Haar and Walsh transforms differ from Fourier and Discrete Cosine transforms primarily in their use of non-sinusoidal basic functions. This makes them distinct in handling nonsmooth or sharply changing data efficiently. The application of Haar and Walsh transforms is often favored in scenarios demanding fast computations and suitable for sharply defined, non-periodic data, unlike the sinusoidal nature favored by Fourier and Discrete Cosine transforms .

The main types of image transforms are orthogonal and non-sinusoidal transforms. Orthogonal transforms use basic functions that are sinusoidal in nature, such as the Fourier Transforms and Discrete Cosine Transforms. Non-sinusoidal transforms, on the other hand, use non-sinusoidal basic functions like Haar Transforms, Walsh Transforms, and Hadamard Transforms. The key difference lies in the mathematical functions used as the basis for these transforms.

You might also like