Image segmentation - general superpixel segmentation & center detection & region growing

Image segmentation toolbox

Build Status codecov Codacy Badge Run Status Coverage Badge CircleCI

Superpixel segmentation with GraphCut regularisation

Image segmentation is widely used as an initial phase of many image processing tasks in computer vision and image analysis. Many recent segmentation methods use superpixels because they reduce the size of the segmentation problem by order of magnitude. Also, features on superpixels are much more robust than features on pixels only. We use spatial regularization on superpixels to make segmented regions more compact. The segmentation pipeline comprises (i) computation of superpixels; (ii) extraction of descriptors such as color and texture; (iii) soft classification, using a standard classifier for supervised learning, or the Gaussian Mixture Model for unsupervised learning; (iv) final segmentation using Graph Cut. We use this segmentation pipeline on real-world applications in medical imaging (see a sample images). We also show that unsupervised segmentation is sufficient for some situations, and provides similar results to those obtained using trained segmentation.


Sample ipython notebooks:


input image segmentation

Borovec J., Svihlik J., Kybic J., Habart D. (2017). Supervised and unsupervised segmentation using superpixels, model estimation, and Graph Cut. In: Journal of Electronic Imaging.

Object centre detection and Ellipse approximation

An image processing pipeline to detect and localize Drosophila egg chambers that consists of the following steps: (i) superpixel-based image segmentation into relevant tissue classes (see above); (ii) detection of egg center candidates using label histograms and ray features; (iii) clustering of center candidates and; (iv) area-based maximum likelihood ellipse model fitting. See our Poster related to this work.

Sample ipython notebooks:


estimated centres ellipse fitting

Borovec J., Kybic J., Nava R. (2017) Detection and Localization of Drosophila Egg Chambers in Microscopy Images. In: Machine Learning in Medical Imaging.

Superpixel Region Growing with Shape prior

Region growing is a classical image segmentation method based on hierarchical region aggregation using local similarity rules. Our proposed approach differs from standard region growing in three essential aspects. First, it works on the level of superpixels instead of pixels, which leads to a substantial speedup. Second, our method uses learned statistical shape properties which encourage growing leading to plausible shapes. In particular, we use ray features to describe the object boundary. Third, our method can segment multiple objects and ensure that the segmentations do not overlap. The problem is represented as energy minimization and is solved either greedily, or iteratively using GraphCuts.

Sample ipython notebooks:


growing RG ellipse fitting

Borovec J., Kybic J., Sugimoto, A. (2017). Region growing using superpixels with learned shape prior. In: Journal of Electronic Imaging.

Installation and configuration

Configure local environment

Create your own local environment, for more see the User Guide, and install dependencies requirements.txt contains list of packages and can be installed as

@duda:~$ cd pyImSegm  
@duda:~/pyImSegm$ virtualenv env
@duda:~/pyImSegm$ source env/bin/activate  
(env)@duda:~/pyImSegm$ pip install -r requirements.txt  
(env)@duda:~/pyImSegm$ python ...

and in the end terminating…

(env)@duda:~/pyImSegm$ deactivate


We have implemented cython version of some functions, especially computing descriptors, which require to compile them before using them

python build_ext --inplace

If loading of compiled descriptors in cython fails, it is automatically swapped to numpy which gives the same results, but it is significantly slower.


The package can be installed via pip from the folder

python install


Short description of our three sets of experiments that together compose single image processing pipeline in this order:

  1. Structure segmentation
  2. Center detection (and ellipse fitting)
  3. Region growing with a shape prior

Annotation tools

We introduce some useful tools for work with image annotation and segmentation.

Structure segmentation

We utilize (un)supervised segmentation according to given training examples or some expectations. vusial debug

The previous two (un)segmentation accept configuration file (JSON) by parameter -cfg with some extra parameters which was not passed in arguments, for instance:

    "slic_size": 35,
    "slic_regul": 0.2,
    "features": {"color_hsv": ["mean", "std", "eng"]},
    "classif": "SVM",
    "nb_classif_search": 150,
    "gc_edge_type": "model",
    "gc_regul": 3.0,
    "run_LOO": false,
    "run_LPO": true,
    "cross_val": 0.1

Center detection and ellipse fitting

In general, the input is a formatted list (CSV file) of input images and annotations. Another option is set -list none and then the list is paired with given paths to images and annotations.

Experiment sequence is following:

  1. We can create the annotation completely manually or use following script which uses annotation of individual objects and create the zones automatically.
     python experiments_ovary_centres/
  2. With zone annotation, we train a classifier for center candidate prediction. The annotation can be a CSV file with annotated centers as points, and the zone of positive examples is set uniformly as the circular neighborhood around these points. Another way (preferable) is to use an annotated image with marked zones for positive, negative and neutral examples.
     python experiments_ovary_centres/ -list none \
         -segs "./data_images/drosophila_ovary_slice/segm/*.png" \
         -imgs "./data_images/drosophila_ovary_slice/image/*.jpg" \
         -centers "./data_images/drosophila_ovary_slice/center_levels/*.png" \
         -out ./results -n ovary
  3. Having trained classifier we perfom center prediction composed from two steps: i. center candidate clustering and candidate clustering.
     python experiments_ovary_centres/ -list none \
         -segs "./data_images/drosophila_ovary_slice/segm/*.png" \
         -imgs "./data_images/drosophila_ovary_slice/image/*.jpg" \
         -centers ./results/detect-centers-train_ovary/classifier_RandForest.pkl \
         -out ./results -n ovary
  4. Assuming you have an expert annotation you can compute static such as missed eggs.
     python experiments_ovary_centres/
  5. This is just cut out clustering in case you want to use different parameters.
     python experiments_ovary_centres/
  6. Matching the ellipses to the user annotation.
     python experiments_ovary_detect/ \
         -info "~/Medical-drosophila/all_ovary_image_info_for_prague.txt" \
         -ells "~/Medical-drosophila/RESULTS/3_ellipse_ransac_crit_params/*.csv" \
         -out ~/Medical-drosophila/RESULTS
  7. Cut eggs by stages and norm to mean size.
     python experiments_ovary_detect/ \
         -info ~/Medical-drosophila/RESULTS/info_ovary_images_ellipses.csv \
         -imgs "~/Medical-drosophila/RESULTS/0_input_images_png/*.png" \
         -out ~/Medical-drosophila/RESULTS/images_cut_ellipse_stages
  8. Rotate (swap) extrated eggs according the larger mount of mass.
     python experiments_ovary_detect/

ellipse fitting

Region growing with a shape prior

In case you do not have estimated object centers, you can use plugins for landmarks import/export for Fiji.

Note: install multi-snake package which is used in multi-method segmentation experiment.

cd libs 
git clone
cd morph-snakes 
pip install -r requirements.txt
python install

Experiment sequence is following:

  1. Estimating shape model from set training images containing single egg annotation.
     python experiments_ovary_detect/  \
         -annot "~/Medical-drosophila/egg_segmentation/mask_2d_slice_complete_ind_egg/*.png" \
         -out ./data_images -nb 15
  2. Run several segmentation techniques on each image.
     python experiments_ovary_detect/  \
         -list ./data_images/drosophila_ovary_slice/list_imgs-segm-center-points.csv \
         -out ./results -n ovary_image --nb_jobs 1 \
         -m ellipse_moments \
            ellipse_ransac_mmt \
            ellipse_ransac_crit \
            GC_pixels-large \
            GC_pixels-shape \
            GC_slic-large \
            GC_slic-shape \
            rg2sp_greedy-mixture \
            rg2sp_GC-mixture \
  3. Evaluate your segmentation ./results to expert annotation.
     python experiments_ovary_detect/ --visual
  4. In the end, cut individual segmented objects comes as minimal bounding box.
     python experiments_ovary_detect/ \
         -annot "./data_images/drosophila_ovary_slice/annot_eggs/*.png" \
         -img "./data_images/drosophila_ovary_slice/segm/*.png" \
         -out ./results/cut_images --padding 50
  5. Finally, performing visualisation of segmentation results toghter with expert annotation.
     python experiments_ovary_detect/



For complete references see BibTex.

  1. Borovec J., Svihlik J., Kybic J., Habart D. (2017). Supervised and unsupervised segmentation using superpixels, model estimation, and Graph Cut. SPIE Journal of Electronic Imaging 26(6), 061610, DOI: 10.1117/1.JEI.26.6.061610.
  2. Borovec J., Kybic J., Nava R. (2017) Detection and Localization of Drosophila Egg Chambers in Microscopy Images. In: Wang Q., Shi Y., Suk HI., Suzuki K. (eds) Machine Learning in Medical Imaging. MLMI 2017. LNCS, vol 10541. Springer, Cham. DOI: 10.1007/978-3-319-67389-9_3.
  3. Borovec J., Kybic J., Sugimoto, A. (2017). Region growing using superpixels with learned shape prior. SPIE Journal of Electronic Imaging 26(6), 061611, DOI: 10.1117/1.JEI.26.6.061611.