hi guys,
the image is attached. please guide me. the average pore diameters were quantified but the reviewer asked me
1. how to determine the average pore wall thickness for each layer?
2 how do we quantify the directionality of the pores in the intermediate?
Hi,
AI image analysis, It uses image processing and machine learning algorithms to analyze and quantify characteristics in images, including the thickness and orientation of structures.
Convolutional neural networks (CNN), a type of deep neural network especially good for computer vision tasks, which can be trained to recognize patterns and characteristics in medical or microscopy images.
Microscopy software with integrated AI, Some advanced microscopy programs already include AI modules that can perform accurate image measurements.
AI is not required here. Basic image processing will do. All the necessary functions are available in OpenCV.
For pore thickness you can first create an edge-image using e.g. Canny edge detection, and then perform distance transform. That will give you the radius of the pores. From there, you can figure out how to derive the quantities you want.
For directionality, you can compute the image brightness derivatives in the x- and y- directions, gx and gy respectively, and then compute the direction of the derivative by doing atan(gy/gx). The pores are perpendicular to the direction of the derivative, so the angle you're looking for (in radians) will be equal to pi/2 - atan(gy/gx).
This can all be achieved if you have at least some knowledge of Python. There are plenty of good tutorials on OpenCV with examples similar to what you need to do here.
Alexey's suggestion about determining the directionality is excellent!
I tried it on image b and image c, see attachment. Plotted are binned dx/dy of all pixels. The spikes at multiples of a half are caused (I think) by the low color-depth of the bitmap: discretization artifacts.
Alexey's suggestion about the pore wall thickness is also good, but only if you have images with much better resolution. The pore walls in the images are just about one pixel thick, so a distance transform will not help you.
I also have my doubts about the usefulness of edge detection in this case, but if you upload better quality images, I can take a look.
- Badges
- Science topic
- Similar topics
- Data Mining
- Classification