How to find center of many objects included in single image

6 ビュー (過去 30 日間)
sana3 sal
sana3 sal 2018 年 5 月 10 日
編集済み: Florian Morsch 2018 年 5 月 17 日
Hello, I need some solution to find the center of the discs in the next image automatically. then i have to draw the profile of the line that crossing these points. I have the attached reference, i really don't understand what they did when telling about the step of "Locating the Central Sagittal Cross-section". So can anyone help me PLEASE :(

回答 (1 件)

Florian Morsch
Florian Morsch 2018 年 5 月 11 日
They used a canny edge detection and then compared the found objects with a reference.
What you want to do is to check for the "disks", use a edge detection ( e.g. canny https://de.mathworks.com/help/images/edge-detection.html?searchHighlight=edge%2520detection&s_tid=doc_srchtitle ) and then use regionprops() ( https://de.mathworks.com/help/images/ref/regionprops.html?s_tid=srchtitle ) to find the center of the found objects. After that you can draw a line from point to point.
  2 件のコメント
sana3 sal
sana3 sal 2018 年 5 月 16 日
Hello there, thank you for your answer, i did the canny edge detection. but when doing the regionprops() to find the objects or their centers. i will find many objects contained in the image, not only the discs! what i really need is to detect the inter-vertebral discs. just like the attached image
Florian Morsch
Florian Morsch 2018 年 5 月 17 日
編集済み: Florian Morsch 2018 年 5 月 17 日
If you check the documentation on regionprops() you will see different properties you can use. Try to figure out which combination suits your case.
For exsample you could take a look on all found objects. The vertebras should be bigger (pixelwise) then other objects, maybe even the bigges. So you could filter regionprops() for only the biggest objects.
Maybe you have to do some image processing first, e.g. if you know that the spine will be on the same position in multiple images you could try to create a mask so you see only a part of the image, then check that part, which will erase some false objects.
EDIT: After i have taken another look on the paper you gave, they find the vertebrae with a canny edge detector and then run a comparison to a predefined model. So they know what the vertebrae looks like and find all objects looking the same found with the canny per correlation.

サインインしてコメントする。

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by