Table of Content

30 November 2015, Volume 33 Issue 6

Previous Issue    Next Issue

Special Issue: Information Security of Multimedia Contents

Tripartite Trade System Based on Secure Watermarking

XIN Xin, QIAN Zhen-xing, ZHANG Xin-peng, DAI Shu

2015, 33(6):  575-584.  doi:10.3969/j.issn.0255-8297.2015.06.001

Asbtract ( 981 )   PDF (3839KB) ( 694 )  

References | Related Articles | Metrics

To ensure security of e-commerce, we propose a tripartite trade system based on secure watermarking. We add a third party (server) in the buyer-seller watermarking protocol. The seller sends his/her own image to the server, and let the server trade with the buyer. The seller first uses homomorphic encryption and spreading to embed the server's watermark in the original image and send the marked image to the server. Having received the marked image, the server uses spread-spectrum dither modulation to embed another watermark to be used for locating suspicious buyers who illegally distribute the image. The server then executes the buyer-seller watermarking protocol based on homomorphic encryption. In the tripartite trade of the e-commerce platform, the proposed system not only prevents both the server and buyer from being framed, but also can trace illegal distribution of any buyer or the server. The system also provides a solution to the secure online transactions for a large number of users.

A Separable Scheme of Reversible Data Hiding in Encrypted Images

TIAN Ye, XIANG Shi-jun

2015, 33(6):  585-594.  doi:10.3969/j.issn.0255-8297.2015.06.002

Asbtract ( 1003 )   PDF (3062KB) ( 859 )  

References | Related Articles | Metrics

In this paper, a separable reversible data hiding method in encrypted images is proposed. The content owner tests an original uncompressed image and generates a location map before embedding. The data-hider embeds additional data into significant bits of the selected pixels according to the location map with an embedding key. At the receiver side, the original image can be perfectly recovered by a context adaptive interpolation algorithm based on spatial correlation of adjacent pixels. If all additional data are embedded into the most significant bits of the selected pixels, quality of the directly decrypted images are poor, and if all data are embedded into the second most significant bits, perfect restoration of the original image cannot be achieved. To resolve these problems, the proposed method can ensure both good quality of the directly decrypted images and error-less restoration. Experimental results show validity and good performance of the method.

Color Image Hashing with Compressive Sensing and Color Vector Angle

LIU Kai, TANG Zhen-jun, ZHANG Xian-quan, YU Chun-qiang

2015, 33(6):  595-603.  doi:10.3969/j.issn.0255-8297.2015.06.003

Asbtract ( 1254 )   PDF (9881KB) ( 663 )  

References | Related Articles | Metrics

This paper presents a hashing method for color images based on compressive sensing and color vector angles. In the preprocessing the input image is first normalized. The normalized image is then converted to a color vector angle matrix, which is further divided into non-overlapping blocks. Compressive sensing is applied to each block, and the mean of measurement vector is used to form the image hash. Experiments show that the proposed method is robust against normal digital operations, has good discrimination capability, and outperforms three existing methods.

Detection of Digital Image Forgery Based on Chromatic Aberration

CHEN Zhu-yi, FANG Zhen

2015, 33(6):  604-614.  doi:10.3969/j.issn.0255-8297.2015.06.004

Asbtract ( 1230 )   PDF (26863KB) ( 545 )  

References | Related Articles | Metrics

This paper proposes a forgery detection method using the characteristics of longitudinal chromatic aberration. For image objects located in similar depth, we extract sharpness features in each color channel based on local phase coherence. Strength and the direction of longitudinal chromatic aberration are estimated from the sharpness difference among different channels. Characteristics of longitudinal chromatic aberration for multiple targets is analyzed, and the inconsistency is exploited to identify image forgery. Experimental results show that the method is effective for composite images.

Image Tampering Recovery Based on Multi-codebook Vector Quantization

JI Ping, QIN Chuan

2015, 33(6):  615-627.  doi:10.3969/j.issn.0255-8297.2015.06.005

Asbtract ( 1363 )   PDF (6724KB) ( 578 )  

References | Related Articles | Metrics

We propose a self-embedding fragile watermarking scheme to detect image tampering and recover original contents. By analyzing edge complexity of all non-overlapping blocks in the original image, the codebooks of vector quantization with different sizes are chosen to compress these blocks. The corresponding indicator and the index value representing the main content of the image block are collected to form the compression codes. They are duplicated to be used as reference-bits. These reference bits are then embedded into other multiple blocks according to secret keys, and authentication-bits are embedded into each block itself. Since the reference-bits based on multiple-codebook vector quantization have high efficiency in representing the image contents, the proposed scheme can achieve good performance of content recovery as compared to the reported methods. This is verified by experimental results.

Multi-focus Image Fusion Based on Texture Features in DCT Domain

TANG Zhen-hua, LIANG Cong, OU Cheng, HUANG Xu-fang, QIN Tuan-fa

2015, 33(6):  628-636.  doi:10.3969/j.issn.0255-8297.2015.06.006

Asbtract ( 1448 )   PDF (13272KB) ( 678 )  

References | Related Articles | Metrics

Exiting multi-focus image fusion methods based on discrete cosine transform (DCT) domain cause blocking effects and artifacts in the fused images. To address this problem, an image fusion algorithm concentrated on texture features in the DCT domain is proposed. The proposed method separates each 8×8 DCT block into several texture regions having various directivity of energy distribution in the source images. Similarity of the frequency spectra for the two texture regions are calculated. The part of the fused image is obtained by selecting the one with more energy, or using the weighted result. Experimental results show that the proposed algorithm can effectively reduce blocking effects and artifacts. Compared with exiting approaches in DCT domain, the proposed algorithm can provide fused images with better subjective quality.

Lossless Data Hiding in Palette Images Based on Mix-ary Codes

LÜ Jin-peng, LIANG Hai-hua, ZHANG Xin-peng

2015, 33(6):  637-643.  doi:10.3969/j.issn.0255-8297.2015.06.007

Asbtract ( 1253 )   PDF (9501KB) ( 545 )  

References | Related Articles | Metrics

This paper proposes a lossless data hiding method in palette images based on mix-ary codes. According to the distributionof indices, the indices are divided into effective indices and redundant indices. In the palette a certain number of redundant indices are mapped to the color of an effective index. Then the indices mapping to the same color stand for different secret digits in order to embed extra data losslessly. A dynamic programming algorithm is used to find the best allocation for redundant indices to achieve the highest embedding capacity. Embedded data subsets carried by different colors are different notational sequences. The data embedding operation does not affect the original image.Experiments showthat themethod has higher embedding capacity than previous methods, especially for medical images.

Reversible Data Hiding in Encrypted Medical Images with Improved Embedding Capacity

JIANG Fei, LI Jun, QIAN Zhen-xing

2015, 33(6):  644-654.  doi:10.3969/j.issn.0255-8297.2015.06.008

Asbtract ( 1050 )   PDF (10937KB) ( 1005 )  

References | Related Articles | Metrics

This paper proposes a scheme of reversible data hiding for encrypted medical images. The proposed scheme contains three parts:a sender, a server and a receiver. The content owner encrypts the original medical image using a stream cipher and uploads the encrypted image to the server. The data-hider on the server hides additional data into the encrypted medical image using a circulating shift method that preserves the size of the encrypted image. The receiver downloads the encrypted image containing additional data, extracts the additional data precisely, and recovers the original medical image losslessly. Compared with some previous schemes, the proposed method has larger embedding capacity.

Region-Selected Image Encryption with Combination of DNA Coding and Chaos

PENG Shan, LONG Min

2015, 33(6):  655-662.  doi:10.3969/j.issn.0255-8297.2015.06.009

Asbtract ( 1216 )   PDF (4247KB) ( 625 )  

References | Related Articles | Metrics

A scheme for encrypting user-defined image region is proposed. The region may be chosen interactively by the user using a mouse or directly entering position parameters. Key streams are generated from a chaotic system. The image is randomly substituted based on DNA encoding. Information about the user-defined region is embedded into the encrypted image with reversible watermarking before transmission. The receiver can directly extract the region information without relying on any additional transmission. Tests and analysis show that the scheme can meet security requirements with good flexibility and practicability.

Improved LSB±k Steganography Based on Decision Surface Optimized by SFLA

OUYANG Chun-juan, LIU Chang-xin, LIU Huan

2015, 33(6):  663-670.  doi:10.3969/j.issn.0255-8297.2015.06.010

Asbtract ( 1025 )   PDF (10636KB) ( 439 )  

References | Related Articles | Metrics

Most improved steganographies based on intelligent optimization cannot realize high-dimensional features optimization simultaneously. To solve the problem, this paper proposes an improved LSB±k steganography (denoted SFLA-LSB±k) based on SFLA to achieve an optimal decision surface. Different from the other improved steganographies that attempt to keep image features unchanged after data embedding as much as possible, the proposed method tries to randomly change feature directions of stego-image in the embedding process optimized by SFLA. Thus, it is difficult to find a decision surface to distinguish cover images from stego-images. Simulation indicates that, with the same embedding capacity, SFLA-LSB±k demonstrates better performance in resisting steganalysis than the traditional LSB±k and the improved LSB±k optimized by PSO. Especially, the AUC value is reduced to 0.5637 when k=1 against steganalysis with 78-dimension features.