Derivation of blur-invariant features using orthogonal Legendre moments

C. Y. Wee, R. Paramesran

Research output: Contribution to journalArticleResearchpeer-review

16 Citations (Scopus)


A novel solution based on the orthogonal Legendre moments to derive a blur-invariant feature set for degraded image recognition systems is proposed. The image is degraded because of the introduction of the point spread function (PSF), which is caused by the imperfect imaging system and the environment. The moment relation between the PSF and discrete-space image has been developed using a discrete, finite-extent 2D convolution model. The derivation of a blur-invariant feature set is made possible because the Legendre central moments are not equally affected by the PSF during the imaging process. Hence, the formulation of a novel technique to derive the blur-invariant feature set using the least-affected Legendre central moments has been proposed. It is made less vulnerable to the additive random noise by setting the order (p+q) of the invariant features to be small. The advantages of the proposed blur-invariant feature set in terms of invariant to PSF, discriminability and noise stability are validated through experiments.

Original languageEnglish
Pages (from-to)66-77
Number of pages12
JournalIET Computer Vision
Issue number2
Publication statusPublished - 2007
Externally publishedYes

Cite this