I am working in medical and nature images processing.
I want to find a popular method to segment texture image.
Could you suggest an approach to do so?
Both the feature extraction and classification.
$\begingroup$
$\endgroup$
3
-
1$\begingroup$ Hi! Do you want to segment or to classify? Anyway, there is no silver bullet. I implemented over 15 textural descriptor algorithms (for an older project) and I can't tell you which one is the best. Just to add some more complexity, the algoritm parameters matters a lot! $\endgroup$– visoftCommented Aug 4, 2014 at 20:59
-
$\begingroup$ @visoft: My task is image segmentation. Now I find that gabor filter is one kind of best method. Which is method that you used $\endgroup$– JohnCommented Aug 5, 2014 at 1:58
-
$\begingroup$ @John, Could you review my answer? $\endgroup$– RoyiCommented Jul 11, 2023 at 16:40
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
You may take one of 2 approaches here:
Classic Machine Learning
- Feature Extraction
You should use methods such as Wavelet Scattering, SIFT and other image features. - Classifier
Probably either Kernel SVM or Boosted Trees.
- Feature Extraction
Deep Learning
If you have many labeled example you may use Deep Learning based approaches. See Image Texture Analysis Using Deep Neural Networks.
In this case the net will do the feature extraction for you.
You should use Transfer Learning and in case you have many samples yet not many of them are labeled you may use Active Learning / Semi Supervised approaches.
Some nice post about the subject Texture Analysis with Deep Learning for improved Computer Vision.