Image segmentation is an open problem, many papers every year on the topic. 

Here's a recent paper in the medical imaging domain that employs the latest state-of-art supervised segmentation techniques:

Jos´e Ignacio Orlando, Matthew Blaschko.

Learning fully-connected CRFs for blood vessel segmentation in retinal images.

Medical Image Computing and Computer Assisted Intervention (MICCAI), Sep 2014

https://hal.inria.fr/file/index/docid/1024226/filename/OrlandoMICCAI2014.pdf

Image segmentation will always be an open

problem because of

- overlapping objects such as bones and

tissue must be separated employing knowledge

about the shape of those objects

- border patterns must be learned "on the fly".

- intelligence might be required to continue

a border line where it is badly illuminated

Regards,

Joachim

There is a recent article published on a review of image carotid segmentation methods. A state of the art image and video segmentation techniques are presented there for the segmentation of medical images of carotid artery.

C.P. Loizou, “A review on ultrasound common carotid artery image and video segmentation techniques”, Med. Biol. Eng. Comput.,accepted.

It can be downloaded from research gate.

Many of the latest segmentation algorithms I saw presented or published are atlas-based segmentations that use registration to an atlas of baseline segmentations to get the segmentation of the new image.

There are also new statistical methods like patch-based segmentation.

But it all depends on what you segment and in what granularity.

1.preprocessing to enhance the image for segmentation

2.Segmentation with the combination of thresholding and optimization technique.

3.Segmentation can also be done by providing the thresholded image  as the initialization.

3.Sometimes segmentation is not required at all.Instead of segmentation the image can be cropped either by considering the overlap or non overlapping of pixel.The window size is selected on trail and error basis.

Answer to "Is image segmentation an open problem?":

Not only in medical image processing, but in any signal processing and classification problem (eg. handwriting, speech, degraded document), segmentation is the only real research problem to be solved. If we solve the segmentation problem adequately, we have very powerful classifiers today to finish the rest of the issues in any recognition problem.

Dear Mohammed Naser,

I highly recommend you to follow Professor H.R. Tizhoosh (from Waterloo university) articles and projects, especially Type-2 Fuzzy Image Segmentation approaches.

It works better than OTSU in handling the uncertainty in noisy images and you wont lose the edges!

Best regards

There is a recent paper in biomedical image segmentation based on machine learning:

Machine Learning Techniques for Biomedical Image Segmentation: An Overview of Technical Aspects and Introduction to State-of-Art Applications , Authors: Hyunseok Seo, Masoud Badiei Khuzani, Varun Vasudevan, Charles Huang, Hongyi Ren, Ruoxiu Xiao, Xiao Jia, Lei Xing , arxiv.org, 2020.