Content-based image retrieval using Legendre chromaticity distribution moments
Pew‐Thian Yap, Raveendran Paramesran · IEE Proceedings - Vision Image and Signal Processing · 2006
It is a well-known fact that the direct storing and comparison of the histogram for the purpose of content-based image retrieval (CBIR) is inefficient in terms of memory space and query processing time. It is shown that the set of Legendre chromaticity distribution moments (LCDM) provides a compact, fixed-length and computation effective representation of the colour contents of an image. Only a small fixed number of compact LCDM features need to be stored to effectively characterise the colour content of an image. The need to store the whole chromaticity histogram is circumvented. Consequently the time involved in database querying is reduced. It is also shown that LCDM can be computed directly from the chromaticity space without first having to evaluate the chromaticity histogram.