This project was concluded in 2017, but the page is still being maintained as an archive of outcomes from the project. See also AAU VBN repository.
08 Dec 2016
We just had a new paper published in
Ultramicroscopy in the January, 2017
issue. This paper details a method for using weights in iterative thresholding
algorithms to enhance the reconstruction of undersampled AFM images. For a
broad range of test images our proposed weighted iterative threshold algorithms
outperform both non-weighted iterative thresholding and l1 based methods.
Abstract
The use of compressed sensing in atomic force microscopy (AFM) can potentially
speed-up image acquisition, lower probe-specimen interaction, or enable super
resolution imaging. The idea in compressed sensing for AFM is to spatially
undersample the specimen, i.e. only acquire a small fraction of the full image
of it, and then use advanced computational techniques to reconstruct the
remaining part of the image whenever this is possible. Our initial experiments
have shown that it is possible to leverage inherent structure in acquired AFM
images to improve image reconstruction. Thus, we have studied structure in the
discrete cosine transform coefficients of typical AFM images. Based on this
study, we propose a generic support structure model that may be used to improve
the quality of the reconstructed AFM images. Furthermore, we propose a
modification to the established iterative thresholding reconstruction
algorithms that enables the use of our proposed structure model in the
reconstruction process. Through a large set of reconstructions, the general
reconstruction capability improvement achievable using our structured model is
shown both quantitatively and qualitatively. Specifically, our experiments
show that our proposed algorithm improves over established iterative
thresholding algorithms by being able to reconstruct AFM images to a comparable
quality using fewer measurements or equivalently obtaining a more detailed
reconstruction for a fixed number of measurements.
Download
04 Nov 2016
Our newest version of the Magni software package was just released on
the 2nd of November. We have not usually mentioned new releases of the
package on this website, but this particular release has some
interesting features we hope some of you find particularly
interesting.
The major new features in this release are approximate message passing
(AMP) and generalised approximate message passing (GAMP) estimation
algorithms for signal reconstruction. These new algorithms can be
found in the magni.cs.reconstruction.amp and
magni.cs.reconstruction.gamp modules, respectively. Note that the
magni.cs sub-package contains algorithms applicable to compressed
sensing (CS) and CS-like reconstruction problems in general - and not
just atomic force microscopy (AFM).
If you are not familiar with the Magni package and are interested in
compressed sensing and/or atomic force microscopy, we invite you to
explore the functionality the package offers. It also contains various
iterative thresholding reconstruction algorithms, dictionary and
measurement matrices for 1D and 2D compressed sensing, various
features for combining this with AFM imaging, and mechanisms for
validating function input and storing meta-data to aid
reproducibility.
The Magni package was designed and developed with a strong focus on
well-tested, -validated and -documented code.
Download
- The package can be found on GitHub where we continually release
new versions: GitHub - release
1.6.0
here.
- The package documentation can be read here:
Magni documentation
- The package can be installed from
PyPI or from
Anaconda.
20 Sep 2016
We had a presentation at the 2016 Scientific Computing with Python (SciPy) conference on validation of function arguments in
Python. This is an attempt at advancing correctness of results by providing an
intuitive function argument validation scheme for Python signal processing
applications. An accompanying paper has now been published in the conference
proceedings. SciPy 2016 was held in Austin, Texas, USA, July 11 - 17, 2016.
Abstract
Python does not have a built-in mechanism to validate the value
of function arguments. This can lead to nonsensical exceptions, unexpected
behaviour, erroneous results and the like. In the present paper, we define
the concept of so-called application-driven data types which place a layer of
abstraction on top of Python data types. With this concept in mind, we discuss
the current argument validation solutions of PyDBC, Traitlets and Numtraits,
MyPy, PyValid, and PyContracts. We find that they share the issue of expressing
the validation scheme in terms of Python objects rather than in terms of the
data they hold. Consequently, we lay out a suggestion for a validation strategy
including what qualifies as a validation scheme, how to create an interface
which promotes both usability and readability, and which Python constructs to
encourage using for validation encapsulation. A reference implementation of the
suggested validation strategy is part of the open-source Python package, Magni
which is thus presented along with a number of examples of the usages of this
package.
Download
20 Sep 2016
We had a presentation at the 2016 Scientific Computing with Python (SciPy) conference on making computational experiments in
Python reproducible. This is an attempt at advancing reproducibility of
computational results by storing metadata describing the computational setup
along with the results. An accompanying paper has now been published in the
conference proceedings. SciPy 2016 was held in Austin, Texas, USA,
July 11 - 17, 2016.
Abstract
Computational methods have become a prime branch of modern
science. Unfortunately, retractions of papers in high-ranked journals due to
erroneous computations as well as a general lack of reproducibility of results
have led to a so-called credibility crisis. The answer from the scientific
community has been an increased focus on implementing reproducible research in
the computational sciences. Researchers and scientists have addressed this
increasingly important problem by proposing best practices as well as making
available tools for aiding in implementing them. We discuss and give an example
of how to implement such best practices using scientific Python packages. Our
focus is on how to store the relevant metadata along with the results of a
computational experiment. We propose the use of JSON and the HDF5 database
and detail a reference implementation in the Magni Python package. Further,
we discuss the focuses and purposes of the broad range of available tools
for making scientific computations reproducible. We pinpoint the particular use
cases that we believe are better solved by storing metadata along with results
the same HDF5 database. Storing metadata along with results is important in
implementing reproducible research and it is readily achievable using scientific
Python packages.
Download
16 Apr 2016
Another paper of ours was presented at the 2016 IEEE
International Symposium on Biomedical Imaging (ISBI) in Prague,
Czech Republic, April 13-16, 2016.
Abstract
The reconstruction quality which can be obtained using
compressive sensing depends on a number of elements. In the
present paper, we establish performance indicators and use these
to model the reconstruction quality of atomic force microscopy
images undersampled with Lissajous sampling patterns. For this
purpose, we consider previously proposed performance
indicators. Furthermore, we propose new performance indicators
based on the relative energy of the subsampled dictionary matrix
atoms. Through extensive simulations, multiple affine models are
evaluated in terms of modified coefficients of determination. The
results show that the proposed performance indicators are highly
correlated with the average reconstruction quality. In
conclusion, the proposed performance indicators can be used to
model reconstruction quality for the given application, and the
proposed model outperforms the previously established model.
Download
21 Jan 2016
We just had a new paper published in IEEE Journal of Selected Topics
in Signal
Processing
in the February, 2016 issue. This is an attempt to create an overview
of some of the basic possibilities in sparse image reconstruction /
inverse problems that can be used to reconstruct images from
undersampled measurements in atomic force microscopy.
Abstract
This paper provides a study of spatial undersampling in atomic force
microscopy (AFM) imaging followed by different image reconstruction
techniques based on sparse approximation as well as interpolation. The
main reasons for using undersampling is that it reduces the path
length and thereby the scanning time as well as the amount of
interaction between the AFM probe and the specimen. It can easily be
applied on conventional AFM hardware. Due to undersampling, it is
necessary to subsequently process the acquired image in order to
reconstruct an approximation of the image. Based on real AFM cell
images, our simulations reveal that using a simple raster scanning
pattern in combination with conventional image interpolation performs
very well. Moreover, this combination enables a reduction by a factor
10 of the scanning time while retaining an average reconstruction
quality around 36 dB PSNR on the tested cell images.
Download
16 Dec 2015
Another paper of ours was presented at the 2015 IEEE Global
Conference on Signal and Information Processing (GlobalSIP) in
Orlando, Florida, USA, December 14-16, 2015.
Abstract
With compressive sensing, the obtainable reconstruction quality
depends on the original signal, the reconstruction algorithm, the
measurement matrix, and the dictionary matrix. The present paper
is concerned with establishing performance indicators and using
these to predict reconstruction quality in atomic force
microscopy applications. For this purpose, we consider the
well-known quantities of coherence and mutual
coherence. Furthermore, we propose a new performance indicator
derived from coherence in order to better model the average
reconstruction quality. Through extensive simulations, affine
models using the performance indicators are evaluated in terms of
modified coefficients of determination. The results show that the
proposed performance indicator yields a better model than both
coherence and mutual coherence do. In conclusion, the proposed
performance indicator can be used to predict reconstruction
quality for the given application, and the affine prediction
model can be improved by including coherence and mutual
coherence.
Download
07 Oct 2014
Our most recent paper was published in the Journal of Open Research Software on October 7, 2014. This is a “software meta-paper”, i.e. a paper describing our software package
Abstract
Magni is an open source Python package that embraces compressed
sensing and Atomic Force Microscopy (AFM) imaging techniques. It
provides AFM-specific functionality for undersampling and
reconstructing images from AFM equipment and thereby accelerating the
acquisition of AFM images. Magni also provides researchers in
compressed sensing with a selection of algorithms for reconstructing
undersampled general images, and offers a consistent and rigorous way
to efficiently evaluate the researchers own developed reconstruction
algorithms in terms of phase transitions. The package also serves as a
convenient platform for researchers in compressed sensing aiming at
obtaining a high degree of reproducibility of their research.
Download
05 Dec 2013
The project’s first paper paper was presented at The 9th International Conference on Signal Image Technology and Internet Based Systems in Kyoto, Japan, December 2-5, 2013.
Abstract
Atomic force microscopy (AFM) is one of the most advanced tools for
high-resolution imaging and manipulation of nanoscale
matter. Unfortunately, standard AFM imaging requires a timescale on
the order of seconds to minutes to acquire an image which makes it
complicated to observe dynamic processes. Moreover, it is often
required to take several images before a relevant observation region
is identified. In this paper we show how to significantly reduce the
image acquisition time by undersampling. The reconstruction of an
undersampled AFM image can be viewed as an inpainting, interpolating
problem, or a special case of compressed sensing. We argue that the
preferred approach depends upon the type of image. Of the methods
proposed for AFM, images containing high frequencies should be
reconstructed using basis pursuit from data collected in a spiral
pattern. Images without too much high frequency content should be
reconstructed using interpolation.
Download
- The original paper is available open access here (IEEExplore),
DOI:
10.1109/SITIS.2013.32.
- The paper is accessible through Aalborg University’s institutional
repository, VBN, here:
VBN