A Two Dimensional Context-Dependent Similarity Measure

Eiichi TANAKA  

IEICE TRANSACTIONS (1976-1990)   Vol.E68   No.10   pp.667-673
Publication Date: 1985/10/25
Online ISSN: 
Print ISSN: 0000-0000
Type of Manuscript: PAPER
Category: Picture Processing

Full Text: PDF(448.1KB)>>
Buy this Article

A picture as a set of pixels is essentially context-dependent. This paper describes the picture matching problem based on the context-dependent edit operations. The edit operations investigated allow substituting pixels of a picture into another pixies, deleting pixels from a picture, and inserting pixels into a picture whose costs determined depending on their contexts. The computational complexity to determine the similarity between pictures with NN pixels is O(N4). An approximate computing method, two separation theorems, isomorphism and molluscoidal isomorphism between pictures are described. This measure will applied to the problem of pattern recognition.