International Journal of Applied Mathematics and Machine LearningAuthor's: Yesen Sun
Pages: [69] - [78]
Received Date: September 31, 2015
Submitted by: Jianqiang Gao.
DOI: http://dx.doi.org/10.18642/ijamml_7100121547
In this paper, we propose a new method of searching for image similar patches after image blocking processing, which combines the gradient information of image and the covariance similarity measure. Gradients convey important visual information and are crucial to scene understanding, covariance similarity measure used to measure the similarity of two variables is very useful, which can capture the image structure and the change of the contrast more effectively in together. Compared with covariance method, the projection method, and face recognition method respectively finding similar patches, a large number of experimental results show that gradient-covariance method is much better than other methods.
image blocking, similarity measure, image gradient, similar patch.
