Catalogue of Artificial Intelligence Techniques
Local Grey-level Operations
Aliases: Image Morphology
Categories: Pattern Recognition and Image Processing , Vision
Author(s): Dave Reynolds
A class of transformations on grey-level image which replace each pixel (in parallel) by some function of its neighbouring pixels. In pointwise operations the new pixel value is independent of the neighbourhood, for example simple grey scale re-mapping such as histogram equalisation. Linear operations (Convolution) are used for filtering (local averaging etc.) and feature detection (such as Edge Detection). Non-linear operations can be developed using local minimum and local maximum functions to replace summation. Such operations can also be regarded as grey level generalisations of binary morphological operations using the Fuzzy Logic based replacement of AND by MIN and OR by MAX. Illustrative operations are `shrink' and `expand' (also called `erode' and `dilate'). In a shrink operation a `true' pixel is changed to `false' if there is a `false' pixel in some defined neighbourhood of it. Similarly for expand. For example, if the neighbourhoods are defined as 3x3 squares then shrink will delete all edge pixels. Some of these types of transformations have been generalised to three-dimensional binary images.
- Serra, J., Image Analysis and Mathematical Morphology
, Academic Press, London , 1982.