Digital holography offers appealing features for 3D imaging applications, overcoming some inherent limitations of current stereoscopic and multiview technologies, with the potential to become the ultimate 3D experience. However, for this vision to become reality, several major technological hurdles need to be overcome. In particular, digital holography requires a very high data rate. Therefore, a key challenge is to define an effective and compact data representation in order to be able to efficiently handle, store, and transmit digital holograms.

In this book, we have addressed this issue and presented an overview of existing techniques for the compression of holographic data. This topic has gained significant interest in recent years, with various directions explored and several innovative solutions proposed. Most of the existing research works have heavily built upon existing compression techniques developed for still images and video sequences, including JPEG and MPEG suites of standards. Nevertheless, digital holograms exhibit rather different features and characteristics, when compared to natural imagery content. Clearly, better statistical modeling of holograms is needed in order to design tailored compression techniques. It is also important to take into account the multidisciplinary nature of the problem of dealing with challenges in optics, signal processing, and compression.

Despite the existing body of published research, more efforts are needed. In particular, we can identify the following open issues which need further investigations.

First, very different settings have been proposed to generate digital holograms, for instance, on-axis versus off-axis or digital recording versus CGH. Moreover, a number of important parameters come into play, including the hologram dimensions, the pixel size, and the wavelength. It results in a wide parameters space which may strongly impact the final hologram and its characteristics. It is therefore paramount to carry out systematic and comprehensive investigations in order to understand better the effect of this wide parameters space. In turn, this also has consequences on the efficiency of subsequent compression techniques.

Second, it is currently impossible to compare results from different publications, as they are using different datasets obtained with different configurations. It is straightforward that common datasets are needed in order to allow researchers to meaningfully compare various approaches. As a first step in this direction, an open database for experimental validations of holographic compression engines was recently proposed in Blinder et al. (2015).¹ It is currently composed of five computer-generated holograms. In addition, it is providing three types of data representation: intensity interference patterns, complex wavefields, and phase-only holograms. However, given the wide parameters space for digital holography as discussed above, further efforts are needed in order to create more comprehensive datasets.

Third, better performance assessment methodologies need to be defined. Fidelity can be measured at two stages: on the compressed hologram or on the reconstructed object, the second way being a priori more meaningful. In addition, most previous research works have used objective quality measures, notably PSNR and SSIM, developed and validated for natural still images. Given that objects reconstructed by digital holography exhibit very different characteristics, this methodology remains questionable until it is methodically validated. In parallel, the visual perception related to the viewing of holograms need to be investigated and understood. To the best of our knowledge, the first subjective experiments to assess the impact of digital holography compression have been carried out in Darakis et al. (2010). More recently, comprehensive subjective studies are reported in Ahar et al. (2015). On a closely related topic, Finke et al. (2015) provides insight in visual perception aspects related to holographic displays. These efforts need to be continued in order to develop and validate objective quality measures which are well-suited in the context of digital holography.

Fourth, most published research publications in this domain have considered simplistic scenes composed of one or very few objects. This is in large part due to limitations of current technology, which is only capable of capturing scenes with small angular size. In order to live up to the expectation of digital holography as the ultimate 3D experience, it is paramount to consider more complex scenes representative of realistic application scenarios.

Last but not least, another key open issue is to understand at which stage of the processing pipeline compression needs to be performed. Most related publications have considered the compression of holograms, as reported in this book. However, CGH allows for another approach, namely to transmit a conventional representation of the 3D scene and to compute the hologram at the display end. This alternative results in two benefits. Firstly, given that a conventional representation of the 3D scene is transmitted, it can be efficiently compressed with traditional 3D video coding schemes. Hence, the development of new compression techniques specifically for holography data is avoided. Secondly, with such a scheme, the representation of the scene can be completely decoupled from the display device, thus offering greater flexibility. Such an approach is proposed in Senoh et al. (2014) for holographic TV. It is based on multiview video coding and depth map coding, and holograms are generated at the receiving end. The authors claim that the proposed scheme uses 1/97,000 of the data rate when compared to the compression and transmission of holograms, while achieving an equivalent subjective quality.

While significant progresses have already been achieved in recent years, by tackling the above open issues, further significant performance gains can be expected in the coming years, thus opening new horizons for digital holography.