102
B. Simon and O. Haeberlé
optical transmission microscopy, gray-level images mix these quantities, which are
here clearly distinguished.
Figure 4.11c, d shows 3D views of the absorption within the pollen (in yellow),
and of the complex index of refraction (refraction and absorption, refraction being
segmented in cyan), respectively. The photopolymer tip, used to handle and rotate
the pollen, is clearly visible on Fig. 4.11d, but not on Fig. 4.11c. The absorption
image (yellow) appears clearly being embedded within the refraction image (cyan),
as shown on the cut through the pollen (Fig. 4.11e).
This interesting property of clearly separating absorption and refraction has, to
the best of our knowledge, not yet been investigated for biological research, while
refraction and absorption images indeed deliver complementary information [67, 83,
86, 146].
Note that other approaches are possible to obtain improved and isotropic resolution
in TDM. A 4-Pi TDM setup, combining transmission and reflection acquisitions, has
been proposed in [81]. A simplified variant has also been investigated [150], cleverly
making use of the mirroring effect to achieve 4-Pi detection using only a single
objective and a special object support. These two techniques have, however, not yet
been experimentally implemented.
4.7 Conclusion and Perspectives
Tomographic diffractive microscopy is able to deliver high-resolution images of unlabeled samples, with various implementations, and attracts the interest of more and
more end-users in various domains. It, however, still suffers from some limitations,
which may contribute to slow-down its adoption.
Because it is not limited by low-level fluorescence signal, it has the potential to
be a fast imaging technique. However, its sequential nature limits the final speed
of acquisition, a single 3D images requiring tens to hundreds of holograms to be
recorded. Also, its computational nature limits final 3D image delivery rate. This
motivates developments to increase speed of data acquisition, as well as to fasten
image computation and rendering [151–161]. Ultra-fast cameras can be used to accelerate acquisitions, fewer images can be taken to compute tomographic images when
requirements in image quality and resolution are less severe, and modern Graphic
Processing Unit (GPU) computation can greatly fasten image reconstruction. So, in
practice, the present bottleneck is in fact the data transfer between fast camera and
fast computer, which still limits imaging acquisition and rendering speed. But progresses in the domain are rapid, as well as in the domain of 3D display systems. We
believe that by combining rapid acquisitions with fast transfer rates and fast GPU
reconstructions, a real-time 3D display of the observed sample images would be
possible [161], which could greatly facilitate adoption of tomographic imaging by
nonspecialists.
Another active domain of research is about simplifying the data acquisition system, which, in its most common form, is based onto an interferometric setup, with
B. Simon and O. Haeberlé
optical transmission microscopy, gray-level images mix these quantities, which are
here clearly distinguished.
Figure 4.11c, d shows 3D views of the absorption within the pollen (in yellow),
and of the complex index of refraction (refraction and absorption, refraction being
segmented in cyan), respectively. The photopolymer tip, used to handle and rotate
the pollen, is clearly visible on Fig. 4.11d, but not on Fig. 4.11c. The absorption
image (yellow) appears clearly being embedded within the refraction image (cyan),
as shown on the cut through the pollen (Fig. 4.11e).
This interesting property of clearly separating absorption and refraction has, to
the best of our knowledge, not yet been investigated for biological research, while
refraction and absorption images indeed deliver complementary information [67, 83,
86, 146].
Note that other approaches are possible to obtain improved and isotropic resolution
in TDM. A 4-Pi TDM setup, combining transmission and reflection acquisitions, has
been proposed in [81]. A simplified variant has also been investigated [150], cleverly
making use of the mirroring effect to achieve 4-Pi detection using only a single
objective and a special object support. These two techniques have, however, not yet
been experimentally implemented.
4.7 Conclusion and Perspectives
Tomographic diffractive microscopy is able to deliver high-resolution images of unlabeled samples, with various implementations, and attracts the interest of more and
more end-users in various domains. It, however, still suffers from some limitations,
which may contribute to slow-down its adoption.
Because it is not limited by low-level fluorescence signal, it has the potential to
be a fast imaging technique. However, its sequential nature limits the final speed
of acquisition, a single 3D images requiring tens to hundreds of holograms to be
recorded. Also, its computational nature limits final 3D image delivery rate. This
motivates developments to increase speed of data acquisition, as well as to fasten
image computation and rendering [151–161]. Ultra-fast cameras can be used to accelerate acquisitions, fewer images can be taken to compute tomographic images when
requirements in image quality and resolution are less severe, and modern Graphic
Processing Unit (GPU) computation can greatly fasten image reconstruction. So, in
practice, the present bottleneck is in fact the data transfer between fast camera and
fast computer, which still limits imaging acquisition and rendering speed. But progresses in the domain are rapid, as well as in the domain of 3D display systems. We
believe that by combining rapid acquisitions with fast transfer rates and fast GPU
reconstructions, a real-time 3D display of the observed sample images would be
possible [161], which could greatly facilitate adoption of tomographic imaging by
nonspecialists.
Another active domain of research is about simplifying the data acquisition system, which, in its most common form, is based onto an interferometric setup, with