Investigation of Geometric Invariance Properties of Hu Moments for Image Processing Applications

Authors : Ethem Sefa Küçükyilmaz, Erhan Akın
Pages : 571-580
Doi:10.24012/dumf.1670505
View : 79 | Download : 88
Publication Date : 2025-09-30
Article Type : Research Paper
Abstract :This research aims at estimating sizes of characters in the Turkish alphabet using techniques in image processing and Hu moments. The field of image processing is an important constituent in numerous applications, with specific examples including character recognition, for instance, in cases of optical character recognition (OCR). In this work, estimation of character \\\"S\\\" in terms of its sizes is performed through analysis of character images using techniques in Hu moments. Hu moments have proven effective in shape and object recognition, and through invariability in reflection, rotation, and scales, have become increasingly useful in a variety of applications. The observations derived during character size estimation involved examination of character \\\"S\\\" in its many sizes, following specific preprocessing protocols. Protocols included techniques in preprocessing, extraction of gradients, and sharpening techniques. After preprocessing, calculation of Hu moments took place in an attempt to evaluate character sizes in each case. Observational information showed that sizes of character \\\"S,\\\" in its many sizes, could accurately be determined. Certain exceptional cases, in cases of uncertain borders of letters, necessitated additional refinements, though. The conclusion of this work identifies the use of Hu moments in character size estimation, in addition to its potential use in character and image processing studies. Conclusively, results validate efficiency and accuracy in proposed estimation of sizes, opening its use to complex forms in its range of applicable forms. In future studies, works could expand through incorporation of sophisticated preprocessing and analysis techniques for improvement in effectiveness of Hu moments.
Keywords : Görüntü İşleme, Hu Momentleri, Boyut Tahmini, Ölçekleme.

ORIGINAL ARTICLE URL