Darakis, Emmanouil and Kowiel, Marcin and Näsänen, Risto and Naughton, Thomas J.
Visually lossless compression of digital hologram sequences.
SPIE Proceedings , 7529.
Digital hologram sequences have great potential for the recording of 3D scenes of moving macroscopic objects as
their numerical reconstruction can yield a range of perspective views of the scene. Digital holograms inherently
have large information content and lossless coding of holographic data is rather inefficient due to the speckled
nature of the interference fringes they contain.
Lossy coding of still holograms and hologram sequences has shown promising results. By definition, lossy
compression introduces errors in the reconstruction. In all of the previous studies, numerical metrics were used
to measure the compression error and through it, the coding quality. Digital hologram reconstructions are highly
speckled and the speckle pattern is very sensitive to data changes. Hence, numerical quality metrics can be
misleading. For example, for low compression ratios, a numerically significant coding error can have visually
negligible effects. Yet, in several cases, it is of high interest to know how much lossy compression can be achieved,
while maintaining the reconstruction quality at visually lossless levels.
Using an experimental threshold estimation method, the staircase algorithm, we determined the highest
compression ratio that was not perceptible to human observers for objects compressed with Dirac and MPEG-
4 compression methods. This level of compression can be regarded as the point below which compression
is perceptually lossless although physically the compression is lossy. It was found that up to 4 to 7.5 fold
compression can be obtained with the above methods without any perceptible change in the appearance of video
||Visually lossless digital hologram coding; Digital hologram visual quality assessment; Digital holographic;
||Faculty of Science and Engineering > Computer Science
||19 Jan 2011 17:28
|Journal or Publication Title:
||SPIE - IS&T
Repository Staff Only(login required)
||Item control page