Page 119 - ITU Journal, ICT Discoveries, Volume 3, No. 1, June 2020 Special issue: The future of video and immersive media

ITU Journal: ICT Discoveries, Vol. 3(1), June 2020

Fig. 11 – Reconstruction of complex-valued hologram Piano8K from two different viewing angles using the Numerical Reconstruction Software for
Holography (NRSH).

plane, using either a pixel-based or angle-based declara- objective metrics is currently in the process of being eval-
tion. The latter allows for declaring not only square but uated. However, it is important to realize that numerical
also rectangular apertures. reconstruction gives rise to speckle noise. This noise is
typically jeopardizing the correct operation of more ad-
NRSH supports two operating modes: batch mode and a
vanced quality metrics. A solution to handle this problem
Graphical User Interface (GUI) mode. The software ac-
is to deploy denoising filters, but yet again these are not
counts for the main reconstruction parameters such as
solely impacting the speckle noise but also intrinsic rele-
wavelength, pixel-pitch, reconstruction distance as well
vant information about the scene [29, 30].
as the dimension and position of the synthetic pupil aper-
ture. Finally, a third approach will deploy application-specific
tools that extract for example refractive index informa-
4.6.3 Objective quality assessment tion or phase profiles from microscopic holograms of
respectively, biological samples or interferometric mea-
Three main strategies are foreseen for quality testing. A surements. It will be studied how these measurements
first approach measures the quality of the holograms in are impacted by the codecs under test.
the hologram plane. This is the easiest approach and ba-
sic measurements made are based on mean squared error
(MSE) and peak signal-to-noise ratio (PSNR). These mea- 4.6.4 Subjective testing protocols
surements report the global quality of the hologram but
do not return any information on how measured errors As mentioned earlier, the availability of holographic dis-
translate into distortions in the reconstructed holograms. plays is still poor. Moreover, testing on such (experimen-
tal) displays is time-consuming and intensive for test sub-
Therefore, to evaluate the reconstruction quality of the jects. However, recent research has demonstrated that
decoded holograms, in a second measurement strategy, light field or regular 2D displays can be used instead and
the hologram is back-propagated to the object plane such still provide sufficient insight in the reconstruction qual-
that visually consumable reconstruction of the hologram ity of compressed holographic content [31].
is being produced. Note that the NRSH software is de-
ployed in this context. Please also note that some coding The strategy followed for 2D display is that a few numer-
solutions operate in the object plane as well, particularly ical reconstructions of the hologram, at different depths
in the context of encoding shallow scenes. and angles, are compared for both the reference and com-
pressed hologram in a Double Stimulus Impairment Scale
The NRSH software will not reconstruct solely one view- (DSIS) [25] side-by-side comparison test.
ing position but typically multiple, focusing the recon-
For light field displays, a reconstruction of the hologram
struction on the foreground, middle and background of
is provided for every supported viewing angle, converting
the scene in addition to multiple viewing angles (left, cen-
tre, right, top, bottom). This allows for assessing the im- the hologram into a light field for a particular depth of fo-
pact of the codecs under test at different reconstruction cus. The test subject is subsequently consuming in an au-
points. If the codecs prioritize e.g. lower frequencies as tostereoscopic fashion a light field reconstruction of the
classical codecs do, depth information will be removed hologram. Also here the evaluation procedure is based on
and reconstructions at larger viewing angles will be poor. a Double Stimulus Impairment Scale (DSIS) [25], though
the reconstructed light fields of the reference and com-
Besides classical MSE, PSNR and SSIM metrics, other met- pressed holograms are displayed sequentially to the test
rics can be deployed as well in this domain. A wide set of subject.

© International Telecommunication Union, 2020 97