Robust Image Identification for Double-Compressed and Resized JPEG Images

Abstract

In the case that images are shared via social networking services (SNS) and cloud photo sharing services (CPSS), it is known that the JPEG images uploaded to the services are often re-compressed and resized by the providers. Because of such a situation, a new image identification scheme for doublecompressed JPEG images having different sizes from that of a singled-compressed one is proposed in this paper. The aim is to detect a single-compressed image that has the same original image as the double-compressed ones, even when the sizes of those compressed images are different. In the proposed scheme, a feature extracted from only DC coefficients in DCT coefficients is used for the identification. The use of the feature allows us not only to robustly avoid errors caused by double-compression but also to perform the identification for different size images. The simulation results demonstrate the effectiveness of the proposed one in terms of the querying performance.

Image Identification of Encrypted JPEG Images for Privacy-Preserving Photo Sharing Services

“…A DC coefficient DC in each block is obtained by using the following equation, where I(x, y) represents a pixel value at the position (x, y) in a block. been reported that the features extracted from DC coefficients are effective for recompression in the conventional schemes [21], [22]. Therefore, the DC coefficients are used for the identification in this paper.…”

“…The proposed scheme was compared with four identification schemes (DC signs-based one [22], sparse codingbased [23], quaternion-based [24] and iterative quantization (ITQ)-based ones [25]). In the schemes [23]- [25], the hamming distances between the hash values of the encrypted images were calculated, and images that had the smallest distance were then chosen as the images generated from an original image, after all images were decompressed.…”

An Image Identification Scheme Of Encrypted Jpeg Images For Privacy-Preserving Photo Sharing Services

“…The range of the DC coefficients is [-1024,1016]. It has been reported that the features extracted from DC coefficients are effective for recompression in the conventional schemes [18,19]. Therefore, the DC coefficients are used for the identification in this paper.…”

“…The proposed scheme was compared with four identification schemes (DC signs-based [19], sparse coding-based [20], quaternion-based [21], and iterative quantization (ITQ)-based ones [22]). In the schemes [20][21][22], the Hamming distances between the hash values of the encrypted images were calculated, and images that had the smallest distance were then chosen as images generated from an original image after all images were decompressed.…”

“…Regarding requirement 3, image retrieval and the identification of encrypted images have never been considered for EtC systems, although image identification and retrieval schemes that are robust against JPEG compression have been proposed for unencrypted images [18][19][20][21][22]. However, the performance of these schemes is degraded in the case of identification between encrypted images and corresponding re-encrypted images, so, to satisfy all requirements, novel image identification methods are required.…”