Connective morphology has been an active area of research for more than two decades. Based on an abstract notion of connectivity, it allows development of perceptual grouping of pixels using different con- nectivity classes. Images are processed based on these perceptual groups, rather than some rigid neighbourhood imposed upon the image in the form of a fixed structuring element. The progress in this field has been threefold: (i) development of a mathematical framework; (ii) develop- ment of fast algorithms, and (iii) application of the methodology in very diverse fields. In this talk I will review these developments, and describe relationships to other image-adaptive methods. I will also discuss the opportunities for use in multi-scale analysis and inclusion of machine learning within connected filters.
19-sep-2017 → 21-sep-2017
20th IAPR International Conference, DGCI 2017: null