A high-performance filter (HPF) is proposed for removing noises in corrupted images, where the local features are adopted to preserve the image local structures. Firstly, the Chebyshev’s theorem and fuzzy mean process are used to adaptively estimate the detection parameters. Secondly, the local statistics theory is used to estimate noisy pixels, which is based on Radon transform. Thirdly, the PSNR is used as the evaluation metric to show the advantage of the HPF, which is compared with latest filters, such as SBF, HPFSM and the classical filter SMF. Extensive experimental results show that the HPF achieves better performances in removing any kinds of noises (impulse, Gaussian, uniform and mixed), the computational complexity of the HPF is 5 to 11 times lower than the other filters.
Robust Image Hashing Based on Local Features for Image Authentication
