Abstract :
In a number of applications, especially those consisting of digital images, searching through large, unstructured databases based on sample sequence is often desirable. This paper evaluates symbolic-based approach for shape description of image elements and shapes. The idea is to identify each shape and to describe the shape as time series function. Symbolic-based algorithm is used to convert time series into one or more symbolic words. The approach was modified to include multiple stages and research showed some interesting conclusion about defects of original approach, the effect of the size of alphabet, symbolic words and general usability of symbolic algorithms for image analysis.
Keywords :
data mining; image matching; image retrieval; image analysis; shape description; shape indexing; shape retrieval; symbolic-based approach; unstructured databases; Aggregates; Approximation algorithms; Clustering algorithms; Data mining; Discrete wavelet transforms; Image sequence analysis; Indexing; Information retrieval; Shape; Vectors; Symbolic aggregation; data mining; image analysis; shape mathcing;
