Singular Value Decomposition Experiments

Experimenting with Singular Value Decomposition

Basic SVD image decomposition

The Jupyter Notebook svd_image_test.ipynb contains experimental code for breaking down a Lena image into factoring matrices and reconstructing it using truncated outer products.

SVD truncation for image denoising

In theory most of the higher sigma values correspond to outer products with higher SNR. Therefore truncating these products from the summation theoretically leads to denoising.