Does anyone know the difference between pattern recognition and image classification?

closed as too broad by Vince, PolyGeo, underdark Jul 5 '14 at 14:21

Please edit the question to limit it to a specific problem with enough detail to identify an adequate answer. Avoid asking multiple distinct questions at once. See the How to Ask page for help clarifying this question. If this question can be reworded to fit the rules in the help center, please edit the question.

  • 3
    A search on the internet provides a huge number of links...what have you looked at and what have you ruled out? This question is quite vague and would benefit from some explanation as to what you don't understand from what you have researched. – user681 Jun 28 '14 at 13:40

those are different things.

Image classification is the process of creating a thematic image where each pixel is assigned a number representing a class (can include the class 'unclassified'). In an aerial image the classes can be soil, vegetation, water etc. image classification algorithms examples are k-means or ISO-DATA.

Pattern recognition is the process of finding things in an image, for example, search for tanks in an aerial/satellite military image or search for a cancerous cells in a medical image or a garden surrounded by buildings or face recognition. usually this will include specific logic about the pattern you want to find.

Most image analysis algorithms are pixel based. However, a major drawback of pixel based algorithms (by definition) is that they do not take contextual information into consideration (context in this case is used to describe the relationship between a pixel/object and its environment).

A possible solution to the problems of pixel based image analysis is to operate in the spatial scale of the objects of interest. Object based image analysis (OBIA) is a science field dedicated to dividing images to meaningful image-objects and evaluating their characteristics in the spatial, spectral and temporal aspects.

Pixel based classification (left) versus object based classification (figures from Aplin & Smith 2008) Pixel based classification (left) versus object based classification (figures from Aplin & Smith 2008)

Image segmentation is a process of aggregating neighboring pixels into objects (regions, segments), such that every object is homogeneous (relative to a certain homogeneous criterion e.g texture), but union of two neighboring objects is not homogeneous. Image segmentation algorithms examples are Watershed, and various edge based, region growing and split and merge algorithms.

this is the place to mention that there is no best image segmentation as it is subjective. give several people to segment an image and you'll have several different segmentation (similar probably but not the same). i think this is the case for allot of image processing algorithms

so to sum it up: pattern recognition algorithms and image classification algorithms have different aim.

  • pattern recognition algorithms can use image classification.
  • image classification can use an image segmentation first and then classify the segments - OBIA.
  • image segmentation can use an image classification for initial segments. (one may think that image segmentation can be done via image classification and then separating connected components of classes, but this is not a 'pure' segmentation because pixels from different places in the images affected each other's segment)
  • Excellent answer @dowi, so Image Classification works on each pixel in turn and Pattern Recognition works on groups of pixels... is that right? – Michael Stimson Jun 29 '14 at 4:36
  • @MichaelMiles-Stimson please see my edit – dowi Jun 29 '14 at 6:00

Not the answer you're looking for? Browse other questions tagged or ask your own question.