Is there a way that I could prove that the pattern of distribution of white pixels in the two attached images are different?

Hello,

In the two attached binary images, I want to find a way to prove that the pattern of white pixels in the two images are different, so is there any way in MATLAB statistics or digital image processing to approach such a problem?

Thanks for your guidance,

[Walter Roberson]

texture analysis perhaps?

... For those times when you are talking about "pattern" as something that might be rotated or translated or scaled/

[Penny13]

Thanks for your kind reply. I havent done texture analysis before, I will definitly look it up,but could you be more specific? I read about spatial analysis and point apttern analysis but I dont know how exactly approach those methods and how to apply them to my problme. In fact I have to group of images and I want to prove that these two groups are different and at the end use the difference for supervised learning.

Thank you for your guidance.

[Image Analyst]

Well you could do standard particle size analysis (area, perimeter, solidity, axes lengths, etc.)

Looks like they've been binarized from gray level images. Can you attach the gray level images that they came from?

One alternative is to just feed the training images into a deep learning network and let the network classifiy them.

[Walter Roberson]

It might be worth measuring fractal dimension.

[Penny13]

Thank you @Image Analyst. I will look into particle size analysis. I also attached gray level images. The other thing is I want to possibly use SVM for training and testing because I want to find the difference between those two group of the images first.

[Penny13]

Thank you @ Walter Roberson. I will look fractal dimention too.

Answer 1

[Image Analyst]

Well there are tons of things you could measure. Area fraction, nearest neighbor distances, area distribution, blob intensities, etc. Literally dozens upon dozens of things.

One approach is to just measure everything you can think of, then use principal components analysis, pca(), to make a model. You could threshold on the first principal component if there are two classes.

Or you could try kmeans() if you know there are two classes.

Or you should try the Classification Learner app to find out which classification/prediction scheme works best given the measurements you have made. Often random forest, though very ad hoc, turns out to be best. Or try several of them and have them "vote" like "adaboost" methods.

But don't discount deep learning - it may end up being the best of them all.

[Penny13]

Thank you so much Image Analyst. Just could you be more specific about PCA? After I measured several properties of the images then how could I use them in PCA?

[Image Analyst]

I attach my PCA demo that uses 3 "measurements" (the red, green, and blue values of an RGB image). And another pca demo I have created (I forget what it's about).

[Penny13]

Thanks a lot. I appreciate your guidance.

[Rodrick]

Hello Penny13. I am writing a Matlab code for a project (Intelligent Tobacco Curing Control Based on Color Recognition). I am using a Raspberry Pi and Pi Camera to capture images of leaves and then the code has to determine the curing stage. The curing stages are identified by the color of the leaf and there are basically 3 main colors(identifiers). There are two parameters(temperature and humidity) which must then be adjusted to required levels unique to each of the 3 stages.

The colors are: 1. Green (Initial Stage) 2. Yellow (Yellowing Stage) 3. Lemonish Colour (Colour fixing stage). A captured process image will be compared with a dataset in Matlab to recognise the curing stage. After recognising the color and stage MATLAB has to then command the Pi to Switch on a control mechanism or end effector, for the prototype it (the end effector) may just be a small Fan to show control of humidity and temperature adjustments will be shown by switiching on of a LED for each temperature level.

So my question is, what are the resources you recommend for me to read to accomplish my goals and have you done something almost similar before ?