Scribd
Upload
5 views

Image_Compression_Presentation

This document outlines a study on image compression techniques using Singular Value Decomposition (SVD) and Fourier Transform, focusing on their efficiency, computational complexity, and quality retention. The study aims to implement and compare these methods based on performance metrics like PSNR, MSE, and compression ratio, while also considering their suitability for various applications. A timeline for the implementation and testing phases is provided, along with references to relevant literature.

Uploaded by

pranavrmenon488
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
5 views

Image_Compression_Presentation

This document outlines a study on image compression techniques using Singular Value Decomposition (SVD) and Fourier Transform, focusing on their efficiency, computational complexity, and quality retention. The study aims to implement and compare these methods based on performance metrics like PSNR, MSE, and compression ratio, while also considering their suitability for various applications. A timeline for the implementation and testing phases is provided, along with references to relevant literature.

Uploaded by

pranavrmenon488
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 7

IMAGE COMPRESSION USING

FOURIER TRANSFORM AND SVD


Group Number :7

Group Members :SANDHIYA ,HARIPRID,PRANAV,SREYAS

Date of Presentation :19-Mar 2025


BASICS LEARNT SO FAR
▪ Singular Value Decomposition (SVD) in Image Compression
▪ Discrete Fourier Transform (DFT) for Images
▪ Fast Fourier Transform (FFT) and Inverse FFT Algorithms
▪ Comparison of Image Compression Methods
▪ Performance Metrics: PSNR, MSE, Compression Ratio
OBJECTIVE
▪ The goal of this study is to analyze and compare different image compression techniques—Singular
Value Decomposition (SVD) and Fourier Transform to evaluate their efficiency in terms of compression
ratio, computational complexity, and quality retention. By understanding how each method transforms
image data, we aim to determine their suitability for various applications, such as imaging, general
image storage, and real-time processing.

Specifically, this study will:


Investigate the mathematical principles behind SVD and Fourier Transform -based compression
techniques.
•Implement and test these techniques to analyze their performance.
•Measure compression efficiency by evaluating factors like Peak Signal-to-Noise Ratio (PSNR), Mean
Squared Error (MSE), and compression ratio.
•Compare the advantages and limitations of each technique in terms of data loss, image reconstruction
accuracy, and computational cost.
•Provide insights into selecting the most appropriate method based on the application requirements,
such as imaging (where lossless quality is critical) or multimedia storage (where higher compression
ratios are preferred).
DETAILED EXPLANATION - SVD
Mathematical Intuition
Image as a Matrix • Large singular values contribute more to
the image’s structure.
• Grayscale images are matrices of pixel
intensities. • Smaller values mostly represent noise or
fine details.
• RGB images have three matrices (one
for each channel: R, G, B). Storage Reduction
• Original storage: m × n values in A , instead
Applying SVD
we store:
• Decomposes matrix A into UΣVᵀ. Uₖ (m × k)
Σₖ (k × k)
• Singular values in Σ capture energy Vₖᵀ (k × n)
(importance).
• Significantly reduces storage when k <<
Compression via Low-Rank min(m, n)
Approximation Trade-off
• Keep only top K singular values. • Lower k → higher compression, but lower
image quality.
• Approximate A using Aₖ = UₖΣₖVₖᵀ.
• Higher k → retains more details, but
requires more storage.
DETAILED EXPLANATION - FFT
•Quantization: Approximate frequency values
•Image as a Signal: Images can be treated as 2D to reduce storage requirements.
signals with spatial and frequency components.
•Reconstruction: The inverse FFT (IFFT)
•Fourier Transform: Converts images from the converts the modified frequency data back to
spatial domain to the frequency domain. an image.

•FFT Efficiency: Reduces DFT computational •JPEG Compression: Uses DCT (related to
complexity from O(N)2 to O(N log N). FFT) for efficient image compression.
•Low vs. High Frequencies: Low frequencies
store most image information; high frequencies •Lossy Nature: Some details are lost, but
capture details/noise. perceptual quality is preserved.

•Compression Strategy: Remove small high- •Trade-off: Higher compression reduces file
frequency coefficients, reducing data without size but increases visual artifacts.
major quality loss.
VACATION PLAN & TIMELINE (APRIL 5 -
MAY 5)
▪ April 5 - April 10: implementation and testing of SVD based compression
▪ April 11 - April 18: Implement and test Fourier Transform based compression
▪ April 19 - April 25: Compare results using PSNR, MSE, and Compression Ratio
▪ April 26 – May 1st: Experiment with improving any of those methods and try
combining them
▪ May 2 - May 5: Finalize documentation and presentation
REFERENCES
▪ P. H. S., S. H. L., and B. M. K. N., "Image Compression Using SVD," International Conference on
Computational Intelligence and Multimedia Applications (ICCIMA), 2007, pp. 143-145.

▪ J. D. Villasenor, "Full-frame compression of tomographic images using the discrete Fourier


transform," IEEE International Conference on Image Processing, 1993, pp. 195-203.

▪ A. T. G. and S. Ramachandran, "Novel algorithms for 2-D FFT and its inverse for image
compression," 2013 International Conference on Signal Processing, Image Processing and Pattern
Recognition (ICSIPR), 2013, pp. 1-4.

You might also like