The SVD-Image-Compression project aims to explore the practical applications of linear algebra, focusing on image processing using Singular Value Decomposition (SVD). The primary goal is to reduce the amount of data necessary to reconstruct an image with minimal loss of quality.
We applied SVD compression to several images, each time varying the k parameter. The k parameter represents the percentage of singular values retained, where a smaller k results in more information loss from the original image. The results, including the space saved as a percentage of the original matrix size, can be found in the images/ folder.
- SVD compression is most effective on larger images when the k parameter is set to values less than 0.2.
- For RGB images, noticeable noise can appear even with higher k values.
Figure 1: Example of an image compressed using SVD with k=0.1.
- Adam Kaniasty
- Igor Kołodziej
- Hubert Kowalski