Image Segmentation using Thresholding

I am currently trying to segment images using thresholding but I don't know what thresholding teechnique should I use for these type of images. I am familiar with Otsu but it does not give me the result that I want that is, only show the pest and do not include the background. Please if you can give me sample codes, it is greatly appreciated. Thank you.

<<

<<

>>

>>

[Rik]

Have a read here and here. It will greatly improve your chances of getting an answer.

Answer 1

[Image Analyst]

Since you have a wide variety of different subjects and colors and textures in your images, I think that deep learning is a possibility. You can use easier, more traditional methods if you narrow down the types of images more, like to only dark bugs on green leaves or something.

[CINDY LASCO]

What I actually posted are images of pests that belongs to different classes however, is there no way for me to get a proper thresholding method to use without narrowing down my images? It is because I am getting each and every image that I can find since unfortunately, I have limited source for image data about banana pests so I am using whatever there is to use for my training data.

[Image Analyst]

Correct. I can't imagine any way that thresholding would get the transparent green pest on a green leaf, and if you did find some way it may only work with that pest on those leaves, not for all the rest of your images. However deep learning would be able to handle it. Who hired you to do this? Don't they have pests and would be able to supply lots of images?

I suggest you start with the easier subgroup, like dark pests on green leaves. With something that easy, yes you can use thresholding. See my Image Segmentation Tutorial to learn how.