High Order Statistic-Zernike Approach for Image Similarity and Face Recognition
الباحث الأول:
Noor Abd Alrazak Shnain, Zahir M Hussain, Mohammed Abdulameer Aljanabi, Song Feng
الباحثين الآخرين:
Noor Abd Alrazak Shnain, Zahir M Hussain, Mohammed Abdulameer Aljanabi, Song Feng
المجلة:
International Journal of Computer Science and Information Security (IJCSIS)
تاريخ النشر:
3 ديسمبر، 2017
مختصر البحث:
Image similarity is the underlying technology in many computer vision applications and is the source of many algorithms used in image processing. Many of images similarity measures have been proposed in the medical image field and computer vision co…
Image similarity is the underlying technology in many computer vision applications and is the source of many algorithms used in image processing. Many of images similarity measures have been proposed in the medical image field and computer vision community. There is no optimal image similarity measure but a set of measures that are appropriate for particular applications. In this paper, we present a highly efficient hybrid measure for image similarity that is based on statistical and momental measures. We propose a similarity measure called the statistical-Zernike measure (SZM), to determine a reliable similarity between any two images including human faces images. This measure combines the best features of high order statistic (HOS) with Zernike moments (ZMs). Simulation results show that the proposed measure outperforms the well-known similarity measures SSIM, FSIM, ZMs and the state-of-art (KSM) by giving more accuracy in detecting similarity even under distortion and more confidence in recognising human faces images under various conditions of facial expression and pose.
