Image Visual Security Assessment Based On HVS Characteristics

With the widespread popularity of cameras,smart phones,cloud storage and other interactive devices,as well as the rapid development of Internet technology,communication technology and multimedia technology,digital images are easier to obtain and disseminate,and have become an essential part of people's lives.Therefore,the security of digital images has become an increasing concern issue.Selective encryption is a widely used method to protect the security of images.With the development of various selective encryption algorithms,it becomes more and more important to evaluate the visual security of selective encrypted images.Visual security assessment(VSA)can evaluate the visual security of selective encrypted images.In the past decades,many methods have been proposed to evaluate the visual security of images,including utilizing image quality assessment(IQA)to evaluate the visual security and VSA methods which were proposed for the security evaluation of encrypted images.However,the existing IQA methods and VSA methods don't take full account of the visual perception characteristics of the selective encrypted images.Therefore,we design the visual security assessment methods for selective encrypted images based on the human visual system(HVS)characteristics.First of all,in this paper,we propose a visual saliency-guided visual security assessment for selective encrypted images.Considering that humans have different levels of interest in different areas of images,the image visual saliency have an important influence on the visual security assessment.Therefore,we propose a visual saliency-guided visual security assessment for selective encrypted images.Considering that human visual perception is very sensitive to the structure and orientation characteristics of selective encrypted images,we extract the structure and orientation feature maps of the original images and the encrypted images,and measure the similarity of the feature maps of the original images and encrypted images,and then generate the structure and orientation similarity maps.Finally,the visual saliency map of the original image is calculated to weight the similarity of the feature maps and obtain the final VSA score.We conduct comparative experiments on two common encryption image quality databases to evaluate the performance of our proposed VSA scores.The experimental results show that the proposed method achieves better performance compared with other state-of-the-art VSA methods.Secondly,in this paper,we propose a visual security assessment based on visual perception characteristics for selective encrypted images.At first,a multi-resolution representation is established to simulate the hierarchical features of human visual system.Then,the visual saliency maps of the original image is calculated to express the levels of human interest in different regions of the images.Then,we extract the structure and texture features of the original images and the encrypted images.Considering that the structure and texture features of images are related with the pixels,and the pixels' correlation makes some details of image pixels hidden.Therefore,it is necessary to remove this correlation,and the naturalness feature of the image can be obtained by removing this correlation.We extract the structure,texture and naturalness features of the images,and weight the feature similarity maps of the original image and the encrypted image by the visual saliency of the original images,and obtain the final VSA score.The experimental results show that the proposed method has better performance compared with other state-of-the-art VSA methods.