PaperlayerView exact revision on arXiv

Source-linked AI summary

Review on Optical Image Hiding and Watermarking Techniques

Shuming Jiao, Changyuan Zhou, Yishi Shi, Wenbin Zou, Xia Li

arXiv:1804.05483v1cs.CV

TL;DR

Watermarking aims to hide an image’s existence, but optical image watermarking had received little survey attention. This review synthesizes optical systems and methods for embedding their outputs in host images, comparing system features and trade-offs.

  • Problem

    Although optical image watermarking had been investigated, little work had surveyed its techniques, while watermarking aims to conceal the original image’s existence rather than merely transform its content.

    Source: §1 · p. 2

  • Method

    The paper reviews optical image-hiding systems and methods for embedding optical-system output into host images.

    Source: §1 · pp. 2–3

  • Results

    The review compares optical systems’ features, advantages, and disadvantages, and covers embedding methods as a second major aspect.

    Source: §1 · p. 3, §2.8 · p. 10, §4 · p. 15

  • Takeaways & Limitations

    Optical information-security approaches generally offer parallel high-speed processing and multidimensional capabilities relative to digital approaches.

    Source: §1 · p. 1

    Cascaded phase-only mask watermarking is difficult to implement in real optical systems because precise mask alignment is challenging.

    Source: §2.4 · p. 7

Abstract

from arXiv · show

Information security is a critical issue in modern society and image watermarking can effectively prevent unauthorized information access. Optical image watermarking techniques generally have advantages of parallel high-speed processing and multi-dimensional capabilities compared with digital approaches. This paper provides a comprehensive review on the research works related to optical image hiding and watermarking techniques conducted in the past decade. The past research works are focused on two major aspects, various optical systems for image hiding and the methods for embedding optical system output into a host image. A summary of the state-of-the-art works is made from these two perspectives.

1. Introduction

Optical watermarking aims to conceal an image within a host, unlike encryption, which transforms it into ciphertext whose existence may remain visible. This section motivates a review of optical approaches and outlines their criteria and general workflow.

1. Introduction

  • Optical information-security approaches are investigated alongside digital approaches and generally offer parallel high-speed processing and multidimensional capabilities.The introduction situates this interest amid growing concerns about unauthorized acquisition, transmission, manipulation, and distribution of digital content.Source: §1 · p. 1
  • A watermark should preserve host-image quality, remain imperceptible and secure, and, when robust, withstand attacks; fragile watermarks instead support integrity checks.Robust watermarks are described as suitable for copyright protection, while fragile ones change when the host is modified.Source: §1 · pp. 1–2
  • Earlier reviews covered optical encryption and digital watermarking, but little work surveyed optical image watermarking; encryption hides content while watermarking aims to hide the original image’s existence.The review frames optical watermarking against encryption, which produces ciphertext that may be known to a third party.Source: §1 · p. 2
  • Optical schemes typically transform the watermark into optical-system output, embed that output in a host image, and retrieve the original watermark optically or opto-digitally.The host may be a photograph or another optical image, such as a digital hologram; embedding can range from weighted addition to adaptive signal embedding.Source: §1 · p. 2, Figure 1 · p. 2
  • The review covers optical watermarking systems and methods for embedding their output into host images.The paper reviews these topics in Sections 2 and 3, respectively, followed by a brief conclusion.Source: §1 · pp. 2–3

2. Various optical systems for image hiding and watermarking

Optical image-hiding research has explored multiple system architectures, including DRPE. DRPE encrypts a watermark into a complex ciphertext for host embedding and retrieval, with reported extensions and implementation constraints.

2. Various optical systems for image hiding and watermarking

  • Optical image hiding has used DRPE, off-axis and phase-shifting holography, cascaded phase-only masks, JTC, ghost imaging, and ptychography.DRPE was proposed by Refregier and Javidi in 1995.Source: §2 · p. 3, §2.1 · p. 3

2.1. Optical watermarking with Double Random Phase Encoding (DRPE)

  • DRPE transforms the hidden image with two random-phase masks into complex ciphertext, which can be weight-added to a host and decrypted to retrieve the watermark.The host contributes a tolerable level of noise to the retrieved watermark; Fig. 2 illustrates the host, watermark, watermarked image, and retrieval.Source: §2.1 · p. 3, Figure 2 · p. 3
  • Quantizing DRPE ciphertext to fewer discrete real values highly compresses its data size while maintaining reasonable recovered-image quality [12,13].The quantized signal is the complex ciphertext generated from the original hidden image.Source: §2.1 · p. 3
  • DRPE extensions use fractional Fourier transform order for extra attack security, support separate retrieval of multiplexed watermarks, and include phase-truncation asymmetric encryption.The ciphertext can be embedded in a conventional host image or a Fresnel hologram.Source: §2.1 · p. 4
  • A third party may blindly detect DRPE-embedded information because lower-order host-image bit planes show stronger randomness, as assessed by a t test.The reviewed work proposes a method to test whether an image contains DRPE-encrypted secret information.Source: §2.1 · p. 4
  • DRPE offers secure hiding and fast processing, but precise random-mask alignment makes real optical implementation difficult, and the masks produce heavy speckle noise in recovered watermarks.Most investigations described are based on numerical simulation only.Source: §2.1 · p. 4

2.2. Optical watermarking with off-axis holography system

Holograms encode scene information across the hologram, which supports robust watermarking. An off-axis holography system can record a watermark as a hologram for embedding in a host image.

2.2. Optical watermarking with off-axis holography system

  • Because hologram information is distributed globally, one portion can contain information about the entire reconstructed image, supporting resistance to cropping, noise, and distortion.A hologram records and reconstructs a 3D scene on a 2D plane using optical diffraction and interference.Source: §2.2 · p. 4
  • Hologram watermarking is described as suitable for robust applications, including embedding a security-token hologram in an ID-card photograph for authentication.Source: §2.2 · p. 4
  • An off-axis Fourier hologram can be recorded by modulating the watermark with a random phase mask, Fourier transforming it, and superposing an off-axis reference wave; the hologram is then weight-added to a host image.The original watermark can be holographically reconstructed directly from the watermarked image; Fig. 3 presents the optical setup.Source: §2.2 · p. 4, Figure 3 · p. 5

2.3. Optical watermarking with phase shifting holography system

Phase-shifting holography encodes watermark information in interferograms that can be combined to reconstruct the hidden image while suppressing host-image interference. The reviewed work also extends the approach to multiplexed and three-dimensional or micro-image watermarking.

2.3. Optical watermarking with phase shifting holography system

  • PSI embeds watermark information by capturing multiple intensity interferograms formed by interfering the watermark's complex light field with a phase-shifted reference wave.The interferograms are embedded into the host image.Source: §2.3 · p. 5, Figure 4 · p. 5
  • Three interferograms can be combined by addition and subtraction to cancel or suppress the host, reconstruct the encrypted field, and recover the watermark.The encrypted field is formed by DRPE and each interferogram is weight-added to a host image.Source: §2.3 · p. 5
  • Multiple watermark images can be multiplexed by superposition and retrieved with some noise, while two hidden images can modulate the complex object's amplitude and phase to avoid crosstalk.The two modulation roles are assigned correspondingly to the two images.Source: §2.3 · pp. 5–6
  • PSI can retrieve watermarks with reduced or no host-image noise, but one watermark requires multiple watermarked host images; dual-channel PSI needs two interferograms instead of three.Compressive sensing can further reduce the data amount.Source: §2.3 · p. 6
  • Hologram-to-hologram watermarking embeds a DRPE-encrypted hidden 3D object's digital hologram into a host object's hologram, with the two objects reconstructed by DRPE and holography, respectively.The embedding and retrieval scheme is shown in Fig. 5.Source: §2.3 · p. 6, Figure 5 · p. 6
  • A micro image can be embedded into a larger-pixel host by converting both to equal-sized holograms through scaled diffraction, then weight-adding one hologram to the other.Source: §2.3 · p. 6

2.4. Watermarking with cascaded phase only mask architecture

Cascaded phase-only-mask systems display a hidden image when the host image and two computed masks are placed in the optical system, rather than embedding the hidden image directly into the host. The approach avoids a separate embedding operation and is non-destructive to the host, but precise optical alignment is difficult.

2.4. Watermarking with cascaded phase only mask architecture

  • The hidden image is displayed by placing the host image and two phase-only masks in the cascaded system, rather than directly embedding the hidden image into the host.For the Fresnel example, POCS generates the masks with the host at the input plane and the hidden image at the imaging plane; Fig. 6 shows the architecture.Source: §2.4 · pp. 6–7, Figure 6 · p. 7
  • The system can form a concealogram, multiplex multiple watermarks for display at different imaging distances, or use a cascaded Fourier-transform system instead of lensless Fresnel propagation.Multiple images can still use two masks generated with a modified iterative algorithm.Source: §2.4 · p. 7
  • In the gyrator-transform variant, Arnold-randomized watermark information is encoded as output-plane amplitude, the host is input-plane amplitude, and Gerchberg–Saxton computes the phase masks.Source: §2.4 · p. 7
  • A hybrid approach uses RSA encryption with cascaded Fourier transforms: CIFT obtains two phase masks, while the encrypted random-intensity image and masks are weight-added or lower-order-bit-replaced into the host.Source: §2.4 · p. 7
  • The architecture needs no additional operation to insert optical output into the host and is non-destructive, but precise phase-mask alignment makes real optical implementation challenging.Source: §2.4 · p. 7

2.5. Watermarking with Joint Transform Correlator (JTC) architecture

The Joint Transform Correlator (JTC) architecture supports both watermark embedding and retrieval. It encrypts the hidden image with a key code and embeds the encrypted information into a Fourier-transformed host hologram; it is relatively easy and fast to implement experimentally, but supports only 2D object images.

2.5. Watermarking with Joint Transform Correlator (JTC) architecture

  • JTC jointly places the hidden image and key code for encryption, adds the encrypted information to the host image's Fourier-transformed hologram, and retrieves it with the correct key.Fig. 7 depicts embedding and retrieval.Source: §2.5 · p. 8, Figure 7 · p. 8
  • JTC is relatively easy to implement experimentally and has fast optical processing, but unlike holography and ptychography it supports 2D rather than 3D object watermarking.Source: §2.5 · p. 8

2.6. Optical watermarking with ghost imaging system

Ghost imaging reconstructs an object image from correlations between object-beam measurements and reference patterns. For watermarking, its single-pixel detector can record 2D or 3D image information where conventional image-sensor arrays do not work, at the cost of long acquisition from repeated illumination.

2.6. Optical watermarking with ghost imaging system

  • Ghost imaging illuminates the target with an object beam and records its intensity using a single-pixel bucket detector, while a reference beam supplies corresponding patterns for computational reconstruction.The image is reconstructed after many pattern illuminations from the measured intensities and reference patterns; Fig. 8 shows a typical setup.Source: §2.6 · p. 8, Figure 8 · p. 8
  • Recorded detector intensities can serve as ciphertext and reference patterns as the key; the output can be embedded in a host image or hidden in another object's ghost-imaging output.Source: §2.6 · p. 9
  • A single-pixel sensor can record 2D or 3D image information where conventional sensor arrays do not work, but repeated reference-pattern illumination makes watermark acquisition take a long time.Source: §2.6 · p. 9

2.7. Watermarking with ptychography

Ptychography records multiple diffraction-intensity patterns to recover an object's complex amplitude, enabling optical watermarking of 3D objects. Its setup is easier to implement than methods requiring precise alignment, but conventional schemes can require many exposures and host images.

2.7. Watermarking with ptychography

  • Ptychography recovers an object's complex amplitude from multiple diffraction-intensity patterns recorded through partially overlapped probes, without a reference wave.Phase retrieval reconstructs the complex information; Fig. 9 illustrates the optical setup.Source: §2.7 · p. 9, Figure 9 · p. 9
  • For 3D-object watermarking, multiple diffraction patterns are weighted into host images, and each depth section can be reconstructed from patterns extracted from those images.Crosstalk between depth sections can yield noise-like recorded patterns, which the review describes as favorable for watermarking.Source: §2.7 · p. 9
  • A conventional ptychography watermarking scheme requires many optical exposures, diffraction patterns, and host images; a proposed single-shot variant records one pattern of tiny diffraction spots.The review also notes that multiple-image hiding with ptychography has been investigated.Source: §2.7 · p. 9
  • Ptychography is easier to implement experimentally than methods such as DRPE that have strict precision requirements.Source: §2.7 · p. 9

2.8. Comparison of various optical systems for image hiding

The review compares optical image-hiding systems by their advantages and drawbacks. It identifies practical constraints in implementation, acquisition speed, and data volume, alongside optical processing capabilities that may be useful for large collections of images and video.

2.8. Comparison of various optical systems for image hiding

  • Table 1 summarizes the pros and cons of the optical image-hiding systems, but the supplied passages do not specify its individual system-by-system entries.Source: §2.8 · p. 10, Table 1 · p. 10
  • Ghost imaging can use a single-pixel sensor, but watermark-image acquisition takes a long time; ptychography also produces many diffraction patterns.Ghost-imaging speed may be improved by designing reference patterns and reconstruction algorithms while ensuring security strength and imaging speed.Source: §2.8 · p. 10
  • Precise alignment under coherent illumination makes some systems, including DRPE and cascaded phase-only masks, hard to implement experimentally; visual-cryptography operations may ease that alignment requirement.The suggested approach relies on approximate rather than very precise alignment of optical components.Source: §2.8 · p. 10
  • Optical systems' parallel high-speed processing and multidimensional capabilities are identified as advantages for future watermarking of many images and video clips.Source: §2.8 · p. 11

3. Methods for embedding optical system output into host image

The section surveys methods for embedding optical-system output into host images, especially holograms, using tailored digital algorithms. Approaches range from weighted addition to adaptive, attack-resistant, binary-image, and fully digital techniques.

3. Methods for embedding optical system output into host image

  • Embedding typically follows optical conversion of the watermark to ciphertext, then uses tailored digital algorithms that account for the output and host-image signal properties.Hologram-in-photograph embedding is the most extensively investigated case described in the review.Source: §3 · p. 11

3.1. Weighted addition in spatial domain and transformed domain

  • Transform-domain hologram embedding can better preserve image quality, robustness, and stability than spatial-domain weighting, while requiring extra digital operations for retrieval.Methods include DCT, DWT, and SVD; the spatial-domain method can support full optical retrieval.Source: §3.1 · pp. 11–12

3.2. Adaptive data embedding into host image

  • Adaptive embedding schemes seek to hide more ciphertext while damaging the host image less, at the cost of higher complexity.Examples include pixel-adaptive bit replacement, block selection and iterative embedding, and contourlet-domain embedding with optimized parameters balancing imperceptibility and robustness.Source: §3.2 · p. 13

3.3 Enhanced resistance to certain attacks

  • Attack-resistant methods target geometric distortions, cropping, print-scan processing, or integrity monitoring by changing the embedded representation or applying correction methods.Examples include inverse log-polar holograms for rotation and scaling resistance, SIFT-based geometric correction, and semi-fragile watermarking that can detect altered image regions while tolerating some transformations.Source: §3.3 · pp. 13–14

3.4. Watermarking in dithering and halftoning

  • Dithering embeds a binary CGH through different matrices selected by its pixel values, while halftoning embeds hologram information by modulating dot patterns or assigning host and watermark information to wave amplitude and phase.An optical correlator and appropriate key function can reveal watermark information in the binary image.Source: §3.4 · p. 14

3.5. Digital watermarking for holograms

  • For digital hologram integrity checks, fragile watermarking can detect small amounts of bit errors; blockwise DCT coefficient quantization is one reported embedding method.Other reviewed approaches use pseudo-random coefficient locations or Hadamard-domain embedding, which has lower computation complexity than DCT and supports varying capacities with hologram resolution.Source: §3.5 · pp. 14–15
  • Other hologram-host methods encode grayscale watermark images as bits and embed them in binary holograms, or place watermark bits in phase-only holograms; the review identifies broader host and ciphertext coverage as future work.The review also includes fully digital watermarking of digital holograms and notes that embedding algorithms for other ciphertext and host-image types merit further investigation.Source: §3.5 · pp. 14–15

4. Conclusion

The paper reviews optical architectures and processing algorithms for image hiding and watermarking proposed over the past decade, comparing systems and summarizing approaches for embedding their outputs into host images.

4. Conclusion

  • The review covers optical systems from DRPE and holography to JTC, ghost imaging, and ptychography, compares their features, and summarizes embedding methods from weighted addition to adaptive schemes.It also discusses future research directions for optical image hiding.Source: §4 · p. 15
Loading 1804.05483v1…