Image Hash Algorithm Research And Implementation

In recent years,the scale of network multimedia date is growing at an exponential speed.With the growing number of digital cameras and smart mobile-phones,digital images become one of the most applied multimedia information in our daily life.You can upload your photos to the Internet as well as get photos you need at anytime and anywhere.While people enjoy the convenience,the security and authenticity of multimedia information need to be taken into account.Image hashing technology,as a promising emerging technology,provides a safe and effective solution for the manage system of digital images.This paper consists of five chapters.Part one is the introduction,which include a detailed description of the topic,the status of domestic and international studies,the purpose and the significance of the research.The second part of this paper describes the concept of image hashing technology,the basic theory,a clear research framework of this project, and makes further studies on the methods of the extraction process and their characteristics in system.The research puts emphasis on the study of image hashing algorithm based on local information.Two new evolutionary algorithms are proposed in part three and part four.This paper presented a novel image hashing algorithm based on salient region detection in part three.This method includes selecting a specific frequency band for the Gauss difference filtering,converting RGB channels of a host image into the intended channels,computing saliency and producing the hash sequence by a new vector quantization compression algorithm.Our experiment demonstrates the effectiveness of the proposed algorithm and shows that the method is robust to various types of attacks such as image zooming,rotating,Gauss blur,histogram equalization and image median filtering and so on.At the same time the algorithm is superior in image tailoring,image mean filtering and median filtering compared with the contrast algorithm.Part four proposed an image hashing method based on the image edge information.This algorithm combines gray scale image related to the image edge information and robust Harris corner points.This method includes extracting image edges by using separated flow-based Difference-of-Gaussians filter,determining robust Harris corner points of the gray scale image,quantizing and compressing the feature vectors into a group of binary sequences.The receiver operating characteristics curve and the correlation coefficients were used to evaluate the performance.It estimated the robustness of the image hashing algorithm by calculating the maximum values,minimum values,averagevalues and standard deviation values of the correlation coefficients.Our experiment data demonstrates the effectiveness of the proposed algorithms and shows that these two methods are robust to various types of attacks even to the rotation operation.The proposed algorithms are compared with other existed methods,and we used the receiver operating characteristics curve as the analytics tool.The results has proved the superiority of these algorithms.At last,the paper summarized the article and analysed the disadvantages of these methods,also put forward future research direction.