Multi-directional Feature Positioning Retrieval Method of Random Encrypted Images

Abstract

Image encryption is an effective means to ensure image information security, but image encryption makes image features hidden, resulting in blurred image positioning features and cannot provide queryable rules. In this paper, the multi-direction feature retrieval method of random encrypted images is studied comprehensively. Multidirectional binary wavelet is used to decompose specific images, and multi-resolution analysis is used to extract multi-direction features of specific images. The method of image location optimization in random encrypted images is used to eliminate the excessive and repeated image features contained in the specific image by image verification, and the probability of image location errors is reduced. The specific image is retrieved by identifying frequent item sets in random encrypted images that are identical to the multidirectional features of a particular image. The results show that the method can locate the random encrypted image effectively. The accuracy of the image location and the average accuracy of the feature points are about 95 % and 97.3 % respectively, and the anti-noise ability is strong. It provides a scientific means for the rapid positioning of efficient images.