View source for Segmentation

{{Infobox
|image        = [[Image:segmentation-picture.png|200px]]
|header1 = Image Segmentation
|header2 = '''Contact:''' [[Fernando Alonso-Fernandez]]
}}


Welcome to this lecture on '''Image Segmentation Techniques'''

(page created in November 2015, last modification November 2016)


In this lecture we will give an overview of algorithms for image segmentation. Segmentation consists of finding groups of pixels that "go together", so that the image is partitioned into regions of connected pixels that are similar according to some property.


[[Image:coins-segmentation.png|450px]]


Segmentation can be seen complementary to edge detection ('''see lecture on edge detection [http://islab.hh.se/mediawiki/Directionality_Analysis here]'''). Edges define boundaries of objects, while segmentation looks for regions occupied by objects. However, not all edges define an object (our object of interest may have many edges inside), and not all objects are defined by clear edges.


We will start with '''simple segmentation''' algorithms, describing basic thresholding algorithm, seed growing, and split and merge algorithms. These are based on the assumption that the objects of interest have clear "uniform" gray level, and they only make use of pixel intensities. For this reason, a more '''complex segmentation''' algorithm is presented, the '''mean-shift algorithm'''. This allows to make use of many local properties at a pixel location, which are then used to find different "modes" or clusters of pixels in the joint histogram of features.


[[Image:segmentation-mean-shift.png|450px]]



Finding "meaningful" regions in images via segmentation have countless uses in Computer Vision, rather than just visualization. It enables higher reasoning levels of analysis leading towards the analysis of objects and interpretation of scenes. 


[[Image:segmentation-goal.png|450px]]


== Teaching Material ==

Get '''slides''' from Google docs (ppt) [https://drive.google.com/open?id=0BzuKeaCE7_FySUNUVERzejI5ZGs here]

[[Image:intro_to_segmentation.png|250px]]


See here a Youtube '''video commenting the slides''' (November 2014): http://youtu.be/ScGmiRO84HE

[[Image:segmentation-youtube.png|350px]]


'''See the lecture''' in Youtube (November 2014): http://youtu.be/CLO_Iaq7CZI


[[Image:video-lecture-segmentation.png|350px]]



== References and sources ==

{| class="wikitable"
|-
|'''R. Klette, “Concise Computer Vision”,  Springer, 2014'''
* Sections 5.1, (basic segmentation), 5.2 (mean-shift) 
* Get [https://drive.google.com/file/d/0B95UheoCtOzocmZjSVYweHpiNjg/view?usp=sharing chapter 5]
* Site of the author (full book not available, but plenty of other resources): http://ccv.wordpress.fos.auckland.ac.nz/
|}

{| class="wikitable"
|-
|'''R. Szeliski, “Computer Vision:  Algorithms and Applications”, Springer 2010'''
* Sections 5.3 (mean-shift)
* The whole chapter in general is interesting, describing several advanced segmentation methods not shown here
* Full book available online: http://szeliski.org/Book/
|}

{| class="wikitable"
|-
|'''M. Shah, “Fundamentals of Computer Vision”, 1997'''
* Section 3 (very basic read about segmentation)
* Full book available online: http://crcv.ucf.edu/courses/CAP6411/book.pdf 
|}