A hybrid image similarity measure based on a new combination of different similarity techniques

Nisreen Ryadh Hamza, Rasha Ail Dihin, Mohammed Hasan Abdulameer

Abstract


Image similarity is the degree of how two images are similar or dissimilar. It computes the similarity degree between the intensity patterns in images. A new image similarity measure named (HFEMM) is proposed in this paper. The HFEMM is composed of two phases. Phase 1, a modified histogram similarity measure (HSSIM) is merged with feature similarity measure (FSIM) to get a new measure called (HFM). In phase 2, the resulted (HFM) is merged with error measure (EMM) in order to get a new similarity measure, which is named (HFEMM). Different kindes of noises for example Gaussian, Uniform, and salt & ppepper noiser are used with the proposed methods. One of the human face databases (AT&T) is used in the experiments and random images are used as well. For the evaluation, the similarity percentage under peakk signal to noise ratio (PSNR) is usedd. To show the effectiveness of the proposed measure, a comparision anong different similar technique such as SSIM, HFM, EMM and HFEMM are considered. The proposed HFEMM achieved higher similarity result when PSNR was low compared to the other methods.

Keywords


feature similarity index measure (FSIM); gaussian noise; histogram similarity measure (HSSIM); salt and pepper noise; SSIM; uniform noise;

Full Text:

PDF


DOI: http://doi.org/10.11591/ijece.v10i2.pp%25p
Total views : 41 times

Refbacks

  • There are currently no refbacks.


Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.