How can I extract features about image texture?

Question

Kotaro 2013-10-31

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/104500-how-can-i-extract-features-about-image-texture

评论： Constantino Carlos Reyes-Aldasoro 2020-6-4

Thank you for reading my question.

I plan to make a code to detect synapses from EM (electron microscopy) images automatically. To this end, first I'd like to detect the characteristic textures around synapse - synapse vesicle.

Could you look at fig.3 on the following link? http://dsresearch.stanford.edu/community/archive_issue_03.html

I think that this is accomplished by extracting the features about image texture and using classifier. But, I don't know how to extract texture features by Matlab.

I appreciate it if you would advise me about this matter.

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Constantino Carlos Reyes-Aldasoro 2020-6-4

Hello. I tried the link and took me to a Page not Found. You could place an image in this feed to look at it. EM is tricky as the high resolution will imply that regions that may look uniform at other resolutions (i.e. a cell) can be different as you are getting more towards the molecular level.

Show me the image and we can continue.

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Constantino Carlos Reyes-Aldasoro 2020-2-3

2
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/104500-how-can-i-extract-features-about-image-texture#answer_413521

Hello

The basic concept of texture is that you will transform your image, which has a certain texture, to a different space, where the texture will be recognised (i.e. the classification, which can be SVM, or any other, even thresholding if there is clear separation of the textures) depending on what texture method (GLCM, spatial filtering, Fourier filtering) you apply.

GLCM is just one of many methods, simple and effective, but not the most powerful. Have a look at this document for a guide to co-occurrence:

https://wrap.warwick.ac.uk/61393/

And the Matlab function is here:

https://www.mathworks.com/help/images/ref/graycomatrix.html

You could also explore filtering in the Fourier domain, and the matlab code is available here:

https://uk.mathworks.com/matlabcentral/fileexchange/67931-multiresolution-texture-segmentation

This recent paper compares different texture segmentations with different algorithms:

https://www.mdpi.com/2076-3417/9/18/3900/htm

Hope this gives you some ideas of how to proceed. Let me know if you need more information, or if solved, please accept the answer.