Light field image encryption based on spatial-angular characteristic

Highlights

• A novel framework of light image encryption based on the internal characteristics of the light field image is proposed.

• The first proposal is to analyze the characteristics of the spatial and angular domains and to encrypt them separately.

• Sufficient experiments have been done to prove the effectiveness and scalability of the proposed scheme for light field image encryption.

Abstract

The 4D light field (LF) can be sliced into the spatial domain and the angular domain. Different from traditional images, LF image can simultaneously capture the direction and position of light propagation. With the development of image encryption technology, the security protection of LF images need to be paid further attention. Existing LF image encryption methods only explore the spatial information of the LF image, but ignore the angular information of LF image. In this paper, we propose a novel scheme for LF image encryption based on the internal spatial-angular characteristic of LF image. Firstly, considering the characteristic representation of the LF, we use the discrete fractional Fourier transform (DFrFT) to process the spatial domain of the LF image. Afterwards, in the angular domain, we employ the difference in viewing angles extracted from multiple angles to compose an epipolar plane image (EPI). The ciphertext EPI obtained by Arnold transform is displayed as an angular domain ciphertext image after reverse reconstruction. Furthermore, weighted linear blending and Chen’s chaotic system are performed on the ciphertext images in the spatial and angular domains to obtain the final ciphertext image. Experimental results demonstrate the effectiveness and applicability of the proposed scheme for LF image encryption.

Introduction

With the rapid development of computer network and multimedia technology, more and more information is transmitted through the network, and information security is facing increasingly severe challenges. As a typical multimedia resource, the digital image is faced with increasingly prominent security problems. Light field (LF) imaging, as emerging imaging technology, can simultaneously capture the intensity and direction of light propagation, and has attracted more and more attention. Meanwhile, LF have been widely developed in depth estimation [1], compression [2], and quality assessment [3] in recent years. The increasing commercialization of LF cameras such as Picam [4] and Lytro [5] has led to their further popularity. However, there are still few works on the security encryption of LF images. The security protection of LF images need further attention.

Encryption is the most direct and basic method to protect digital image information security. At present, image encryption technologies can be roughly divided into chaotic system [6], DNA sequences [7], compressive sensing [8], and optical image encryption [9]. These types of methods have their own advantages and disadvantages in terms of encryption performance. In [10], Ismail et al. presented a novel lossless image encryption algorithm based on edge detection and generalized chaotic maps for key generation. Wang et al. designed a visually secure encryption scheme by using the parallel compressive sensing (PCS) counter mode and embedding technique [11]. Optical encryption techniques can use the amplitude, phase, and wavelength of light to encode images, which have high speed in parallel processing of 2-D information.

Refregier and Javidi firstly adopted two random phase plates to design double random phase encoding (DRPE) system [12]. The system takes two identical Fourier lenses as the main body, and their distance is the focal length f. The DRPE system has the distinct advantage of parallel processing of two-dimensional complex data and high speed, but its disadvantage is that the degree of freedom of key space is small. Based on the Fresnel diffraction theory, Situ et al. proposed an image encryption method that can achieve efficient encryption [13], but the method occupied substantial computing resources and storage resources. Follow Refregier’s work, Unnikrishnan et al. introduced the fractional Fourier transform (FrFT) theory to design an image encryption scheme, where the lens parameters are used as part of the key to expand the key space [14]. In [15], the authors applied digital holography technology to double random phase encoding, and employ the Joint Transform Correlator (JTC) to build a security system that can be used for video storage and transmission. Zhong et al. proposed a novel approach using discrete multiple-parameter FrFT for image encryption based on interference [16].

Image encryption methods based on optics have been continuously proposed, and various theories have been introduced into the field of image encryption. However, these methods mainly considered the encryption for traditional images. At the same time, there are still little research on LF image encryption scheme. You et al. proposed a micro-lens to record the spatial and directional information for 3-D image encryption [17]. Wen et al. focused on color LF image encryption based on DNA sequence and chaotic system [18]. The authors reported a novel optical encryption strategy that utilizes highly scattered wavefront of LF to encrypt the plaintext and exploits a scattering medium as the unique physical key in Liu et al. [19]. In summary, these methods only explore the spatial information of the LF image and do not fully consider the angular information of LF image.

Different from the existing LF image encryption methods, this paper simultaneously considers the internal spatial-angular characteristic of the LF image, and propose a novel encryption scheme for LF image. Specifically, the discrete fractional Fourier transform (DFrFT) is adopted to extract specific spatial features of LF image. Then, the epipolar plane image (EPI) is generated from multiple views of the LF image. The ciphertext EPI obtained by Arnold transform is displayed as an angular domain ciphertext image after reverse reconstruction. Finally, weighted linear blending and Chen’s chaotic system are performed on the ciphertext images in the spatial and angular domains to obtain the final ciphertext image. The experimental results demonstrate the effectiveness and scalability of the proposed scheme for LF image encryption.

The rest of this paper is organized as follows. Section 2 introduces the related work. Section 3 describes the proposed LF image encryption scheme. Experimental results and analysis are presented in Section 4. Section 5 concludes this paper and discusses future work.