Elsevier

Expert Systems with Applications

Volume 40, Issue 17, 1 December 2013, Pages 6918-6927
Expert Systems with Applications

Increasing image compression rate using steganography

https://doi.org/10.1016/j.eswa.2013.06.008Get rights and content

Highlights

  • Two new image compression algorithms based on DCT-JPEG and DWT-JPEG, called Stego-JPEG (DCT) and Stego-JPEG (DWT), are investigated.

  • The key idea is to compress a block of an image using lossy JPEG, and then hide the resulting bits into subsequent blocks.

  • Increasing compression ratio is improved with high quality for both Stego-JPEG (DCT) and Stego-JPEG (DWT).

  • For color images, Stego-JPEG (DCT) and Stego-JPEG (DWT) have more compression ratio than grayscale also decrease of PSNR is better than grayscale.

Abstract

The goal of image compression is to remove the redundancies for minimizing the number of bits required to represent an image while steganography works by embedding the secret data in redundancies of the image in invisibility manner. Our focus in this paper is the improvement of image compression through steganography. Even if the purposes of digital steganography and data compression are by definition contradictory, we use these techniques jointly to compress an image. Hence, two schemes exploring this idea are suggested. The first scheme combines a steganographic algorithm with the baseline DCT-based JPEG, while the second one uses this steganographic algorithm with the DWT-based JPEG. In this study data compression is performed twice. First, we take advantage of energy compaction using JPEG to reduce redundant data. Second, we embed some bit blocks within its subsequent blocks of the same image with steganography. The embedded bits not only increase file size of the compressed image, but also decrease the file size further more. Experimental results show for this promising technique to have wide potential in image coding.

Introduction

With the advent of the Internet and the need for digital right management systems, steganography has received a particular interest and wide range of applications, especially when it is used in conjunction with cryptography. When these techniques are combined, the secret data is doubly sheltered; first it is encrypted and then embedded within the target support. There are a number of usual and unusual applications of steganography. An example of its usual application is watermarking, which is a replication of a logo or text so that the origin of the target document can be authenticated.

One possible unusual application of steganography is image compression, which is the focus of our paper. In fact, the objectives of digital steganography and data compression are by definition antagonist. However, if steganography process adds extra data within the target support for authentication purpose, compression attempts to remove redundant data to reduce the original file size. To this end, two image compression algorithms exploring this idea are investigated. The first one is based on the baseline DCT-based JPEG, while the second uses the DWT-based JPEG. The baseline JPEG and DWT-based version of JPEG are still widely used for compression of still images available in the Web and produced by digital cameras. To decrease the original file size, a steganographic scheme is integrated within the compression encoding process. Precisely, after the division of the target image into a set of blocks using JPEG, some blocks are embedded into their subsequent blocks of the same image. That is, compression is performed in two steps. First, the conventional standard JPEG either DCT or DWT is used. And second, by means of steganography which embedded some bits-blocks within its subsequent blocks of the same image. The embedded blocks do not increase the file size of the compressed image, but as they are taken from and hidden within the image itself, the file size will be further decreased. The full explanation of our algorithm is presented in Section 3.

Research works related to data embedding based on compression has already been proposed in Swanson et al., 1997, Campisi et al., 2002. In Swanson et al. (1997), the authors use the discrete wavelet transform to split the original image into two parts, the low-pass image (called host image), and the high-pass image (called the residual image). The residual image is, first, coded using a modified version of the embedded zero-tree wavelet coder (EZW) and, then embedded into the host image. The embedding process is based on a linear projection, quantization and perturbation in the DCT domain (2 bits in each 8 × 8 blocks). Unfortunately, the suggested method in Swanson et al. (1997) is not clear and the embedding techniques details are not shown. Moreover the authors limit their work on an image (i.e. Lena) and they do not show the image quality after embedding. In Campisi et al. (2002), an image is, first, presented in its luminance and chrominance components in wavelet domain. Then, the subsampled chrominance is embedded into the luminance component. The embedding regions are subbands where the human visual system has less sensitivity to them. We note that the embedding process in this work is performed without data hiding; rather, the authors used data replacing. Additionally, this method was applied only with unconventional coding (i.e., SPHIT).

In this paper, we introduce a novel algorithm Stego-JPEG (DWT) as an extension to previous work Stego-JPEG (DCT) (Jafari, Ziou, & Mammeri, 2011) to apply steganography in DWT domain. Moreover, we intend to extend our previous framework for color images which provides high compression gain with high quality.

This paper is structured as follows. In Section 2, we present some fundamentals related to JPEG and data hiding in JPEG. In Section 3, the suggested compression scheme is explained. We illustrate our results by a set of simulations in Section 4. Finally, we summarize and present some future directions in Section 5.

Section snippets

Background on data hiding and compression

The necessary background paper is presented in this section. Sections 2.1 briefly describes the lossy JPEG compression scheme based on DCT and DWT, while Section 2.2 presents some background on data hiding in JPEG.

Compressive image hiding scheme

Recall that steganography based compression algorithms must satisfy invisibility, payload, robustness and file size requirements. The embedded data is invisible if a human subject with normal vision is not able to distinguish media that contain hidden data from those that do not. The payload refers to the number of hidden bits while satisfying the invisibility requirement. The embedded data is robust if it can be detected after non intentional modification such as lossy compression. Finally,

Experimental results

Having described the two image compression schemes, we present in this section the evaluation methodology used in this work as well as the simulation parameters.

The compression performance of the schemes under consideration is assessed using both the compression ratio and the quality of compression. Note that the criterion of image quality comparison is the resemblance between original and reconstructed images. For the comparison between images, we employed the peak signal-to-noise ratio (PSNR)

Conclusion

We showed how steganography can be used efficiently to increase the compression ratio of JPEG based on DCT and DWT. Even if the purposes of steganography and compression are antagonist, we illustrated how these techniques can be used jointly to compress and to hide data within the same image. Hence, a novel compression scheme based on JPEG and steganography is explored. The key idea behind our work is to compress the target block of an image using JPEG, and then hide the resulting bits into

References (12)

  • P. Campisi et al.

    Compressive data hiding: An unconventional approach for improved color image coding

    EURASIP Journal on Applied Signal Processing

    (2002)
  • B. Chen et al.

    Quantization index modulation: A class of provably good methods for digital watermarking and information embedding

    IEEE Transaction on Information Theory

    (2001)
  • C. Christopoulos et al.

    The JPEG2000 still image coding system: an overview

    IEEE Transaction on Consumer Electronics

    (2000)
  • R. de Queiroz et al.

    Wavelet transforms in a JPEG-like image coder

    IEEE Transactions on Circuits and Systems for Video Technology

    (1997)
  • R.C. Gonzalez et al.

    Digital image processing

    (2008)
  • Hore, A., & Ziou, D. (2010). Image quality metric: PSNR vs. SSIM. In 20th International conference on pattern...
There are more references available in the full text version of this article.

Cited by (26)

  • Interval type-2 fuzzy logic system based similarity evaluation for image steganography

    2020, Heliyon
    Citation Excerpt :

    The image used for embedding is called the cover image, and the output image after embedding is called the stego image. The overall objective of steganography – to hide the data so that it is indiscernible – can be achieved by ensuring a few things: the protected data inside the cover object must be transparently imperceptible; the method of embedding itself must be able to withstand various attacks; the stego object should have a high embedding capacity for hiding data bits; and the method used should be such that it can resist any damaging interference (Jafari et al., 2013; Jero et al., 2016; Kanan and Nazeri, 2014; Li et al., 2010; Subhedar and Mankar, 2014; Wu and Tsai, 2003). In literature, there are several existing methods for steganography.

  • A new transform domain steganography based on modified logistic chaotic map for color images

    2017, Journal of Information Security and Applications
    Citation Excerpt :

    Statistical results show that algorithm had high resistance against χ2 and RS attacks [30]. In [31] Reza Jafari, et al.are introduced a transform domain steganography. The proposed method uses DWT in embedding and extracting process.

  • Medical data sheet in safe havens - A tri-layer cryptic solution

    2015, Computers in Biology and Medicine
    Citation Excerpt :

    Additionally, authors have adapted the Walsh–Hadamard transform (WHT), binary discrete Hartley transform (BDHT), DCT, and hybrid diagonal block-set transform (FHDBT) techniques for image encryption applications. To further the complexity, authors have developed image encryption algorithms based on compression techniques [26]. Under the compression schemes, the Huffman, multiple Huffman table, run-length and dictionary algorithms were employed to generate cryptic images.

  • A novel image steganography scheme with high embedding capacity and tunable visual image quality based on a genetic algorithm

    2014, Expert Systems with Applications
    Citation Excerpt :

    The term payload is utilized to describe the size of the secret message that can be embedded in a specific image. There are different steganography approaches including spatial domain (direct manipulation of image intensities) (Carvajal-Gamez, Gallegos-Funes, & Rosales-Silva, 2013; Chan & Cheng, 2004; Chen, Chang, & Le, 2010; Ioannidou, Halkidis, & Stephanides, 2012; Naor & Shamir, 1995; Sajedi & Jamzad, 2010; Shamir, 1979; Wu & Tsai, 2003; Yang, Cheng-Hsing et al., 2008), frequency domain (manipulates the image indirectly through various transforms like DFT, DCT, DWT, etc.) (Barni & Mauro, 1999; Chen, 2008; Chu et al., 2004; Jafari, Ziou, & Rashidi, 2013; Liu & Qiu, 2002; Noda, Niimi, & Kawaguchi, 2006) and the compression (substitution) domain (Chang, 2007; Chang, Nguyen, & Lin, 2011; Chang, Tai, & Lin, 2006; Chang, Wu, & Hu, 2007; Chung, Shen, & Chang, 2001; Yang, 2011). Each approach has different specifications.

View all citing articles on Scopus
View full text