How can I implement auto threshold into my code rather than defining the threshold for each image?

Hi everyone. I would like to implement auto-thresholding into my Matlab code as I don't want to define manually for each image. For example, I define my threshold like; binaryImage = myinput <0.7; Is there any way to find the best threshold automatically and use that threshold for images I have.

[Ameer Hamza]

It will be possible most probably. But first, you need to define best threshold. How do you measure whether a specific threshold is best or not?

[engineer]

I find the best threshold by trial and error. The best threshold for me is that the background pixel is fully black (0) and the object's pixel is fully white (0).

Answer 1

[Ameer Hamza]

See adaptthresh(). This will estimate a threshold value without the need to manually specifying it.

[engineer]

I tried, but it did not give me a good result, unfortunately.

[Ameer Hamza]

You can also try some other functions: imbinarize(), otsuthresh(), graythresh() etc.

You can try this FEX submission: https://www.mathworks.com/matlabcentral/fileexchange/8647-local-adaptive-thresholding.

You can also refer to the following resources

https://blogs.mathworks.com/steve/2016/07/25/adaptive-thresholding-for-binarization/

https://www.mathworks.com/matlabcentral/answers/179474-adaptive-threshold-for-segmentation

[engineer]

Thank you very much, I used otsu method and it worked. However, I still have an issue. Because in my image, the object is separated from the background and its pixel is white. However, there is some non-visible white pixel on the image as well. How can I get rid of those white pixels? My calculation is based on the white pixels on the objects. Therefore, other white pixels affect my calculation in a negative way.

[Ameer Hamza]

Use bwareaopen() to remove patches of white pixels which are less than some defined pixel numbers.

[engineer]

thank you very much.

[Ameer Hamza]

You are welcome.