← All projects

Signals & imaging · Thresholding medical images

Thresholding medical images

When should one threshold become several?

Project account · methods and observations from my study and research notes.

01 / The question

When should one threshold become several?

This project compared thresholding methods on DICOM/CT images and investigated how noise changes the separation between intensity classes.

Image → representation → decision
Conceptual illustration of the approach

02 / The intuition

A mechanism worth testing.

A threshold is a decision rule tied to an image distribution. A method that works on one histogram may fail after acquisition conditions change.

03 / The work

Inside the method.

Explore each part of the approach.

01Read the image

DICOM image arrays and intensity histograms provided the starting point.

02Compare decision rules

Iterative global thresholds, Otsu, multi-Otsu and adaptive methods were explored.

03Add perturbations

Noise experiments and mask comparisons examined sensitivity to changing image statistics.

04 / The observations

What emerged.

The notes describe comparative thresholding and noise analysis across several methods.

This account is based on recorded project notes. Original reports, figures and datasets are not embedded here.

05 / The limits

Where the evidence stops.

An intensity class is not automatically an anatomical structure. Ground-truth definitions and the original masks are required to interpret segmentation quality.

The academic foundations

Where this work connects.

Programming & computational methodsSignal & image processing
Explore another projectBrain-image segmentation ↗