Restoration of embedded image from corrupted stego image
Introduction
Data hiding technology embeds secret messages into cover signals such as image, video, audio, graphics, and text [1]. It does not significantly change the perceptual quality of cover signal and then makes unintentional observers unaware of the existence of embedded messages. Thus, it finds applications in copyright protection, content authentication, image forensics, data binding, covert communication, and so on [2]. This paper focuses on image-based data hiding, where cover signal and secret message are both images. The image-based data hiding has many practical applications. For example, to safely transmit a military map from one place to another, one can embed the military map into a cover image to produce a stego image, and then send the stego image by common channels, such as the Internet. As digital images are easily disturbed by impulse noise during the transmission, secret image extracted from the stego image are often corrupted. To ensure that the extracted image has good visual quality, some sophisticated methods for image restoration are needed. One possible approach is to exploit filtering methods for noise removal.
Many filtering methods have been reported in the literature. Alajlan et al. [3] propose a variation of the peak-and-valley filter based on a recursive minimum-maximum method. This method preserves constant and edge areas even under high impulsive noise contamination. Luo [4] presents an efficient algorithm consisting of two steps: impulse noise detection and impulse noise cancellation. This algorithm can preserve image details and does not need previous training. In [5], Yuan et al. use a difference-type noise detector and a cost function-type filter to design a filtering algorithm. The algorithm only requires one iteration independent on people's experiences. In another study [6], Kanga and Wang give a modified switching median filter (MSWM) by adding one more noise detector. This method can detect those noise pixels whose values are close to their neighbors, and then significantly improves the visual quality of noisy images. Chen and Lien [7] design a detector to find noisy pixels and exploit an edge-preserving filter to reconstruct their intensity values. Previous training is also not needed in the algorithm. In [8], Petrović et al. employ genetic programming to remove universal impulse noise, where the genetic programming is used as a supervised learning algorithm for building two noise detectors. This filter does not have any parameters, but requires time-consuming training. In another work, Ibrahim et al. [9] propose a simple filter method for impulsive noise removal. It combines an adaptive median filter with a switching median filter, and does not need threshold parameter. Recently, Wang and Wu [10] present a filtering algorithm based on the minimum absolute value of four convolutions achieved by one-dimensional Laplacian operators. This algorithm can not only preserve image details, but also has a fast computation time. In [11], Zhang uses adaptive center-weighted median filter to identify corrupted pixels and restores them by using a median filter based iterative method. As iteration is needed, computational time is not satisfactory. In another study, Yu et al. [12] remove impulse noise by using a nonmonotone adaptive gradient method (NAGM). The NAGM is a globally convergent and low-complexity method. The above algorithms have shown good performance in noise removal. However, if the stego image is contaminated by impulse noise, some extracted bits from the stego image will be inevitably changed, leading to poor quality of the extracted embedded image. In this case, filtering methods can effectively improve visual quality of stego image, but improvement on the extracted embedded image is limited. So they are ineffective in restoring embedded images from the corrupted stego images.
In this work, we propose a robust method for restoring embedded images with good visual quality from corrupted stego images. The rest of this paper is organized as follows. Section 2 describes the proposed method and Section 3 presents the experimental results. Conclusions are made in Section 4.
Section snippets
Restoration of embedded image
As shown in Fig. 1, our method consists of three steps. The first step is to mark the pixel bits of the embedded image by identifying the changed pixels of the stego image. The second step is to extract the embedded image from the stego image by using data extraction method, which is related to the data embedding method. As our aim is to restore the embedded image, we choose the well-known least significant bit (LSB) substitution [13] from diverse data hiding algorithms [13], [14], [15], [16]
Optimum n value determination
To determine optimum n value of the Eq. (3), we use different images as cover images, embed different secret images to produce stego images, and add impulse noises with different densities into the stego images. In experiments, the used threshold T is 100 and the n values ranging from 1 to 11 are all tested. For each secret image, we extract the embedded images from the corrupted stego images, and calculate peak signal-to-noise ratios (PSNRs) between the original and the extracted images under
Conclusions
In this paper, we have proposed a method for restoring embedded images from stego images contaminated by impulse noises. The rule of impulse noise detection [4] is exploited to identify corrupted pixels of stego image and then the extracted bits of embedded image are marked. Corrupted pixels of the embedded image are finally corrected by adjusting their unreliable bits. As pixel correction is unrelated to noise type, the proposed method is suitable for image restoration from other noise
Acknowledgments
This work was partially supported by the Natural Science Foundation of China (61165009, 60963008), the Guangxi Natural Science Foundation (2011GXNSFD018026, 0832104), the Project of the Education Administration of Guangxi (200911MS55), the Scientific Research and Technological Development Program of Guangxi (10123005–8), and the Scientific Research Foundation of Guangxi Normal University for Doctor Programs. The authors would like to thank the anonymous referees for their valuable comments and
References (16)
- et al.
Reversible image hiding scheme using predictive coding and histogram shifting
Signal Processing
(2009) - et al.
Detail preserving impulsive noise removal
Signal Processing: Image Communication
(2004) An efficient algorithm for the removal of impulse noise from corrupted images
International Journal of Electronics and Communications
(2007)- et al.
Impulse noise removal by the difference-type noise detector and the cost function-type filter
Signal Processing
(2007) - et al.
A new impulse detection and filtering method for removal of wide range impulse noises
Pattern Recognition
(2009) An efficient median filter based method for removing random-valued impulse noise
Digital Signal Processing
(2010)- et al.
Impulse noise removal by a nonmonotone adaptive gradient method
Signal Processing
(2010) - et al.
Hiding data in images by simple LSB substitution
Pattern Recognition
(2004)
Cited by (4)
Secret Image Restoration With Convex Hull and Elite Opposition-Based Learning Strategy
2023, IEEE Signal Processing LettersHigh capacity data hiding based on interpolated image
2017, Multimedia Tools and ApplicationsResearch on self-recovery digital image authentication algorithm
2013, Beijing Jiaotong Daxue Xuebao/Journal of Beijing Jiaotong UniversityA reversible data hiding scheme for JPEG images
2013, ICIC Express Letters