Authors: YongHwan Lee & Tony Storey
This project focuses on removing Poisson noise from medical images using the Expectation Maximization (EM) algorithm. The EM algorithm is applied for poisson noise removal.
Find more details from the report: PDF
where:
-
$Y$ represents the observation matrix. -
$\theta$ is the true coefficient that we aim to estimate. -
$A \in \mathbb{R}^{i \times j}$ is the body model matrix.
Each pixel models the absorption coefficient.
| *--------------------* |
| y3---\ | | | | |
| ---/ | p1 | p2 | p3 | |
- *--------------------* -
| y2---\ | | | | |
| ---/ | p4 | p5 | p6 | |
- *--------------------* -
| y1---\ | | | | |
| ---/ | p7 | p8 | p9 | |
| *--------------------* |
-------|------|-------
y9 y10 y11
|| || ||
\/ \/ \/
*--------------------*
| | | |
| p1 | p2 | p3 |
*--------------------*
| | | |
| p4 | p5 | p6 |
*--------------------*
| | | |
| p7 | p8 | p9 |
*--------------------*
-------|------|-------
To implement the code, follow these steps:
- Clone the repository, which includes the
Med5.m
file. - Run the
Med5.m
file to complete the EM algorithm and estimate the body model matrix coefficients. - The implementation will also plot Mean Squared Error (MSE) graphs, comparing the Cramer-Rao Lower Bound (CRLB) under the fixed body model matrix.
By following these steps, you can effectively execute the algorithm and visualize the results.