Scribd
Upload
3 views

Image Processing and Analysis

Uploaded by

22951a6755
Copyright
© © All Rights Reserved
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
3 views

Image Processing and Analysis

Uploaded by

22951a6755
Copyright
© © All Rights Reserved
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 3

Image Processing and Analysis-AAT2

Name: Nagarakanti Kiran Course: ACDC08


RollNo:22951a6755 AAT-2

1. Basic Relationships Between Pixels


a. Pixels in an image interact based on their spatial arrangement and intensity values.
These relationships determine the structure and content of the image. The four
fundamental relationships are:

b. Neighboring Pixels: Pixels adjacent to a given pixel (in 4-neighborhood, 8-


neighborhood, or diagonal configuration) influence its properties.

c. Connectivity: Pixels are considered connected if they share similar properties, such
as intensity, and are part of the same neighborhood.

d. Edge Relations: Defined by abrupt changes in intensity, edges establish boundaries


between regions of differing characteristics.

e. Region Membership: A group of pixels sharing a consistent property forms a region,


often representing an object or uniform area in the image.

2. Basis Function for Restoration Model


a. To derive a restoration model:

b. Degradation Model: Assume an image g(z,y)g(x, y) is formed as:

c. g(x,y)=f(x,y)∗h(x,y)+n(x,y)g(x, y) = f(x, y) \\\ast h(x, y) + n(x, y)


d. where f(z,y)f(z, y) is the original image, h(z,y)h(z, y) is the degradation function, and
n(x,y)n(z, y) is noise.

e. Fourier Domain Analysis: Transform into the frequency domain:

f. G(u,v)=F(z,v)⋅H(z,v)+N(z,v)G(z, v) = F(z, v) \\cdot H(z, v) + N(z, v)

g. Restoration: Estimate F(u,v)F(u, v) using:

h. F(z,v)=G(z,v)H(u,v)F(u, v) = \\frac{G(u, v)}{H(u, v)}


i. This assumes H(z,v)≠0H(u, v) \\neq 0 and can be refined by regularization techniques
to handle noise.

3. Butterworth Filters

a. Low-Pass Filter: Smoothens the image by reducing high-frequency components:

b. H(u,v)=11+(D(u,v)D0)2nH(u, v) = \frac{1}{1 + \\left(\frac{D(u, v)}{D_0}\right)^{2n}}


c. where D(u,v)D(u, v) is the distance from the origin, D0D_0 is the cutoff frequency,
and nn controls the sharpness.
d. High-Pass Filter: Enhances details by retaining high-frequency components:

e. H(x,v)=11+(D0D(x,v))2nH(x, v) =\\ \rac{1}{1 + \ft(\frac{D_0}{D(x, v)}\ri)^{2n}}

4. Homomorphic Filtering
a. Homomorphic filtering enhances images by treating illumination and reflectance
separately.

b. Model: An image f(x,z)f(x, z) can be decomposed as:

c. f(x,y)=i(x)⋅r(x)f(x, y) = i(x,) \cdot r(x, y)


d. where i(x,y)i(x, y) is illumination and r(x,z)r(x, z) is reflectance.

e. Logarithm Transformation: Converts the product to addition:

f. ln⁡f(x,y)=ln⁡i(x,z)+ln⁡r(x,z)\\ln f(x, y) = \\ln i(x, z) + \\ln r(x, y)

g. Processing Steps:

i. Transform to frequency domain.

ii. Apply a high-pass filter to enhance r(x,y)r(x, y).

iii. Transform back to spatial domain and exponentiate.

h. Block Diagram:

i. Input Image → Logarithm → Fourier Transform → Filter → Inverse Fourier →


Exponentiation → Enhanced Image

j. Image Degradation and Restoration Process

k. Degradation Model: Captures how an image becomes distorted through blurring,


noise, or environmental factors.

5. Restoration Techniques:

i. Inverse Filtering: Reverses blurring.

ii. Wiener Filtering: Balances deblurring and noise reduction.

l. Noise Filters:

m. Mean Filter: Reduces random noise by averaging pixel values.

n. Median Filter: Replaces a pixel with the median of its neighbors, effective against
salt-and-pepper noise.

o. Gaussian Filter: Applies a Gaussian function to smooth the image while preserving
edges.

6.Noise Probability Density Functions (PDFs)

p. Gaussian Noise: Bell-shaped distribution, common in natural systems. p(z)=12πσe−


(z−μ)22σ2p(z) = \\frac{1}{\\sqrt{2\\pi}\\sigma} e^^{{{-\frac{(z-\mu)^2}{2\sigma^2}} }}
Salt-and-Pepper Noise: Random black-and-white pixel corruption.

q. Uniform Noise: Constant probability within a range.

r. Plots: Gaussian: Smooth curve. Salt-and-Pepper: Discrete spikes. Uniform: Flat.

7.Image Visualization on Rigid Body

s. Rigid body transformations involve rotation, translation, or scaling without altering


the image's structure. A visualization demonstrates the effect of these
transformations using a reference grid or object, ensuring alignment and
consistency.

8.Feature Transformation

t. Feature invariance ensures that transformations like rotation or scaling do not alter
the detected attributes. This property is separable for certain algorithms, aiding in
object recognition.

9.Statistical Shape Analysis

u. Shape Standards: Defines models for consistent object representation.

v. Classification Steps:

i. Data preprocessing.

ii. Extracting shape features.

iii. Applying statistical models for class prediction.

10.Feature-Based Registration

w. Involves aligning images by matching key features (e.g., edges or corners). Example:
Aligning satellite images by overlaying matching landmarks.

You might also like