Image segmentation: Difference between revisions

Revision as of 06:02, 3 April 2025

Latest revision as of 04:48, 11 August 2025

(7 intermediate revisions by 5 users not shown)

Line 8:

| title=2003 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2003. Proceedings. |

chapter=On region merging: The statistical soundness of fast sorting, with applications |

publisher=IEEE | year=2003 | volume=2 | doi=10.1109/CVPR.2003.1211447 | pages=II:19–26 | isbn=0-7695-1900-8 }}</ref> such as [[color]], [[luminous intensity|intensity]], or [[Image texture|texture]]. Adjacent regions are significantly different with respect to the same characteristic(s).<ref name="computervision" /> When applied to a stack of images, typical in [[medical imaging]], the resulting contours after image segmentation can be used to create [[3D reconstruction]]s with the help of geometry reconstruction algorithms like [[marching cubes]].<ref>Zachow, Stefan, Michael Zilske, and Hans-Christian Hege. "[https://opus4.kobv.de/opus4-zib/files/1044/ZR_07_41.pdf 3D reconstruction of individual anatomy from medical image data: Segmentation and geometry processing]{{Dead link|date=August 2025 |bot=InternetArchiveBot |fix-attempted=yes }}." (2007).</ref>

== Applications ==

Line 17:

* [[Content-based image retrieval]]<ref>Belongie, Serge, et al. "[http://people.eecs.berkeley.edu/~malik/papers/blobworld98.pdf Color-and texture-based image segmentation using EM and its application to content-based image retrieval]." Sixth International Conference on Computer Vision (IEEE Cat. No. 98CH36271). IEEE, 1998.</ref>

* [[Machine vision]]

* [[Medical imaging]],<ref>{{cite journal | last1 = Pham | first1 = Dzung L. | last2 = Xu | first2 = Chenyang | last3 = Prince | first3 = Jerry L. | year = 2000 | title = Current Methods in Medical Image Segmentation | journal = Annual Review of Biomedical Engineering | volume = 2 | pages = 315–337 | pmid = 11701515 | doi = 10.1146/annurev.bioeng.2.1.315 }}</ref><ref>{{cite journal | last1 = Forghani| first1 = M. | last2 = Forouzanfar | first2 = M.| last3 = Teshnehlab| first3 = M. | year = 2010 | title = Parameter optimization of improved fuzzy c-means clustering algorithm for brain MR image segmentation | journal = Engineering Applications of Artificial Intelligence | volume = 23 | issue = 2 | pages = 160–168 | doi = 10.1016/j.engappai.2009.10.002 }}</ref> and imaging studies in biomedical research, including [[volume rendering|volume rendered]] images from [[CT scan|computed tomography]], [[magnetic resonance imaging]], as well as volume electron microscopy techniques such as FIB-SEM.<ref>{{Cite journal |last1=Reznikov |first1=Natalie |last2=Buss |first2=Dan J. |last3=Provencher |first3=Benjamin |last4=McKee |first4=Marc D. |last5=Piché |first5=Nicolas |date=October 2020 |title=Deep learning for 3D imaging and image analysis in biomineralization research |url=http://dx.doi.org/10.1016/j.jsb.2020.107598 |journal=Journal of Structural Biology |volume=212 |issue=1 |pages=107598 |doi=10.1016/j.jsb.2020.107598 |pmid=32783967 |s2cid=221126896 |issn=1047-8477|url-access=subscription }}</ref>

** Locate tumors and other pathologies<ref>{{cite journal | url=https://link.springer.com/article/10.1007/s11548-013-0922-7 | doi=10.1007/s11548-013-0922-7 | title=Brain tumor detection and segmentation in a CRF (Conditional random fields) framework with pixel-pairwise affinity and superpixel-level features | year=2014 | last1=Wu | first1=Wei | last2=Chen | first2=Albert Y. C. | last3=Zhao | first3=Liang | last4=Corso | first4=Jason J. | journal=International Journal of Computer Assisted Radiology and Surgery | volume=9 | issue=2 | pages=241–253 | pmid=23860630 | s2cid=13474403 | url-access=subscription }}</ref><ref>E. B. George and M. Karnan (2012): "[http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.411.7411&rep=rep1&type=pdf MR Brain image segmentation using Bacteria Foraging Optimization Algorithm]", ''International Journal of Engineering and Technology'', Vol. 4.</ref>

** Measure tissue volumes<ref>{{Cite journal |last1=Ye |first1=Run Zhou |last2=Noll |first2=Christophe |last3=Richard |first3=Gabriel |last4=Lepage |first4=Martin |last5=Turcotte |first5=Éric E. |last6=Carpentier |first6=André C. |date=February 2022 |title=DeepImageTranslator: A free, user-friendly graphical interface for image translation using deep-learning and its applications in 3D CT image analysis |journal=SLAS Technology |volume=27 |issue=1 |pages=76–84 |doi=10.1016/j.slast.2021.10.014 |pmid=35058205 |issn=2472-6303|doi-access=free }}</ref><ref>{{Cite journal |last1=Ye |first1=En Zhou |last2=Ye |first2=En Hui |last3=Bouthillier |first3=Maxime |last4=Ye |first4=Run Zhou |date=18 February 2022 |title=DeepImageTranslator V2: analysis of multimodal medical images using semantic segmentation maps generated through deep learning |language=en |biorxiv=10.1101/2021.10.12.464160v2 |doi=10.1101/2021.10.12.464160 |s2cid=239012446|doi-access=free }}</ref>

** Diagnosis, study of anatomical structure<ref>{{cite journal|last1=Kamalakannan|first1=Sridharan|last2=Gururajan|first2=Arunkumar|last3=Sari-Sarraf|first3=Hamed|last4=Rodney|first4=Long|last5=Antani|first5=Sameer|title=Double-Edge Detection of Radiographic Lumbar Vertebrae Images Using Pressurized Open DGVF Snakes|journal=IEEE Transactions on Biomedical Engineering|date=17 February 2010|volume=57|issue=6|pages=1325–1334|doi=10.1109/tbme.2010.2040082|pmid=20172792|s2cid=12766600}}</ref>

** Surgery planning

Line 213:

== Graph partitioning methods ==

[[Graph (data structure)|Graph]] partitioning methods are an effective tools for image segmentation since they model the impact of pixel neighborhoods on a given cluster of pixels or pixel, under the assumption of homogeneity in images. In these methods, the image is modeled as a weighted, [[undirected graph]]. Usually a pixel or a group of pixels are associated with [[Vertex (graph theory)|nodes]] and [[Glossary of graph theory#Basics|edge]] weights define the (dis)similarity between the neighborhood pixels. The graph (image) is then partitioned according to a criterion designed to model "good" clusters. Each partition of the nodes (pixels) output from these algorithms are considered an object segment in the image; see [[Segmentation-based object categorization]]. Some popular algorithms of this category are normalized cuts,<ref>Jianbo Shi and [[Jitendra Malik]] (2000): [https://www.cs.cmu.edu/~jshi/papers/pami_ncut.pdf "Normalized Cuts and Image Segmentation"], ''IEEE Transactions on Pattern Analysis and Machine Intelligence'', pp 888–905, Vol. 22, No. 8</ref> [[random walker (computer vision)|random walker]],<ref>Leo Grady (2006): [http://vision.cse.psu.edu/people/chenpingY/paper/grady2006random.pdf "Random Walks for Image Segmentation"], ''IEEE Transactions on Pattern Analysis and Machine Intelligence'', pp. 1768–1783, Vol. 28, No. 11</ref> minimum cut,<ref>Z. Wu and R. Leahy (1993): [ftp://sipi.usc.edu/pub/leahy/pdfs/MAP93.pdf "An optimal graph theoretic approach to data clustering: Theory and its application to image segmentation"]{{Deaddead link|date=January 2020May 2025|bot=InternetArchiveBot medic}}{{cbignore|fix-attemptedbot=yes medic}}, ''IEEE Transactions on Pattern Analysis and Machine Intelligence'', pp. 1101–1113, Vol. 15, No. 11</ref> isoperimetric partitioning,<ref>Leo Grady and Eric L. Schwartz (2006): [http://www.cns.bu.edu/~lgrady/grady2006isoperimetric.pdf "Isoperimetric Graph Partitioning for Image Segmentation"] {{Webarchive|url=https://web.archive.org/web/20110719090404/http://www.cns.bu.edu/~lgrady/grady2006isoperimetric.pdf |date=19 July 2011 }}, ''IEEE Transactions on Pattern Analysis and Machine Intelligence'', pp. 469–475, Vol. 28, No. 3</ref> [[minimum spanning tree-based segmentation]],<ref>C. T. Zahn (1971): [http://web.cse.msu.edu/~cse802/Papers/zahn.pdf "Graph-theoretical methods for detecting and describing gestalt clusters"], ''IEEE Transactions on Computers'', pp. 68–86, Vol. 20, No. 1</ref> and [[segmentation-based object categorization]].

=== Markov random fields ===

Line 352:

[[Pulse-coupled networks|Pulse-coupled neural networks (PCNNs)]] are neural models proposed by modeling a cat's visual cortex and developed for high-performance [[biomimetic]] [[image processing]]. In 1989, Reinhard Eckhorn introduced a neural model to emulate the mechanism of a cat's visual cortex. The Eckhorn model provided a simple and effective tool for studying the visual cortex of small mammals, and was soon recognized as having significant application potential in image processing. In 1994, the Eckhorn model was adapted to be an image processing algorithm by John L. Johnson, who termed this algorithm Pulse-Coupled Neural Network.<ref>{{cite journal|last1=Johnson|first1=John L.|date=September 1994|title=Pulse-coupled neural nets: translation, rotation, scale, distortion, and intensity signal invariance for images|doi=10.1364/AO.33.006239|pmid=20936043|publisher=OSA|volume=33|journal=Applied Optics|number=26|pages=6239–6253|bibcode=1994ApOpt..33.6239J}}</ref> Over the past decade, PCNNs have been utilized for a variety of image processing applications, including: image segmentation, feature generation, face extraction, motion detection, region growing, noise reduction, and so on. A PCNN is a two-dimensional neural network. Each neuron in the network corresponds to one pixel in an input image, receiving its corresponding pixel's color information (e.g. intensity) as an external stimulus. Each neuron also connects with its neighboring neurons, receiving local stimuli from them. The external and local stimuli are combined in an internal activation system, which accumulates the stimuli until it exceeds a dynamic threshold, resulting in a pulse output. Through iterative computation, PCNN neurons produce temporal series of pulse outputs. The temporal series of pulse outputs contain information of input images and can be utilized for various image processing applications, such as image segmentation and feature generation. Compared with conventional image processing means, PCNNs have several significant merits, including robustness against noise, independence of geometric variations in input patterns, capability of bridging minor intensity variations in input patterns, etc.

[[U-Net]]In is a2015, [[convolutional neural network|convolutional neural networks]] reached state of the art in semantic segmentation.<ref>{{Cite conference|last=Long |first=Jonathan |last2=Shelhamer |first2=Evan |last3=Darrell |first3=Trevor |date=2015 |title=Fully Convolutional Networks for Semantic Segmentation |url=https://openaccess.thecvf.com/content_cvpr_2015/html/Long_Fully_Convolutional_Networks_2015_CVPR_paper.html |conference=Proceedings of the IEEE conference on computer vision and pattern recognition|pages=3431–3440}}</ref> [[U-Net]] is an architecture which takes as input an image and outputs a label for each pixel.<ref>{{cite arXiv|last1=Ronneberger|first1=Olaf|last2=Fischer|first2=Philipp|last3=Brox|first3=Thomas|title=U-Net: Convolutional Networks for Biomedical Image Segmentation|eprint=1505.04597|date=2015|class=cs.CV}}</ref> U-Net initially was developed to detect cell boundaries in biomedical images. U-Net follows classical [[autoencoder]] architecture, as such it contains two sub-structures. The encoder structure follows the traditional stack of convolutional and max pooling layers to increase the receptive field as it goes through the layers. It is used to capture the context in the image. The decoder structure utilizes transposed convolution layers for upsampling so that the end dimensions are close to that of the input image. Skip connections are placed between convolution and transposed convolution layers of the same shape in order to preserve details that would have been lost otherwise.

In addition to pixel-level semantic segmentation tasks which assign a given category to each pixel, modern segmentation applications include instance-level semantic segmentation tasks in which each individual in a given category must be uniquely identified, as well as panoptic segmentation tasks which combines these two tasks to provide a more complete scene segmentation.<ref name="Panoptic Segmentation"/>