More Website Templates at TemplateMonster.com!

Current Research Lines

Quatitative Susceptibility Mapping in the Brain

Enhancement of the QSM inverse problem formulation, with a more accurate noise and forward model. This research is part of my Ph.D. thesis.

Total Generalized Variation with a Poisson Noise model

Noise reduction and deconvolution of images with a TGV regularization and a Poisson data noise model.

Automatic Image Registration of Non-stellar Astronomical Images

Feature detection and characterization of planetary, lunar and solar images for image registration.

Quality indexing of Non-stellar Astronomical Images

Study of quality estimator using Machine Learning and Pattern Recognition algorithms for the design of a quality sorting algorithm.

Research Papers and Conference Presentations

"Multiscale gradient domain compression for astronomical high dynamic range imaging".
Milovic C, Conejero J, Tejos C. 2014

Abstract:
Visualization of high dynamic range images requires compression of the data to be properly displayed on media with more limited dynamic ranges. Astronomical images pose a difficult challenge for dynamic range compression algorithms, due to the nature of the imaged objects, and to the lack of a reflectance-illumination model based on spatial frequencies. As a result, most of the algorithms commonly used for daylight HDR compression fail in achieving an optimal visualization of astronomical targets. We propose an extended multiscale algorithm based on compression of the dynamic range in the gradient domain. Our algorithm effectively compresses the dynamic range, enhances local contrast, and avoids noise amplification. This is achieved with a multiscale representation of the image, and the use of luminance information. Our results show a significantly improved visualization of astronomical images, compared to the standard gradient domain compression, as well as more robustness to noise and better artifact suppression.

Paper accepted in the Journal of Imaging Science.

"Calcium (Ca2+) waves data calibration and analysis using image processing techniques".
Milovic C, Oses C, et al. 2013

Abstract:
Calcium (Ca2+) propagates within tissues serving as an important information carrier. In particular, cilia beat frequency in oviduct cells is partially regulated by Ca2+ changes. Thus, measuring the calcium density and characterizing the traveling wave plays a key role in understanding biological phenomena. However, current methods to measure propagation velocities and other wave characteristics involve several manual or time-consuming procedures. This limits the amount of information that can be extracted, and the statistical quality of the analysis.
Our work provides a framework based on image processing procedures that enables a fast, automatic and robust characterization of data from two-filter fluorescence Ca2+ experiments. We calculate the mean velocity of the wave-front, and use theoretical models to extract meaningful parameters like wave amplitude, decay rate and time of excitation.
Measurements done by different operators showed a high degree of reproducibility. This framework is also extended to a single filter fluorescence experiments, allowing higher sampling rates, and thus an increased accuracy in velocity measurements.

You can access the full text here BMC BioInformatics, 14, 1

"Quantitative Susceptibility Map Reconstruction via a Total Generalized Variation Regularization".
Felipe Yañez, Audrey Fan, Berkin Bilgic, Carlos Milovic, Elfar Adalsteinsson , Pablo Irarrazaval. 2013

Proceedings of the 2013 International Workshop on Pattern Recognition in Neuroimaging, USA, Philadelphia, 2013, 4

"Leptin and adiponectin have opposite effect on ciliary activity, calcium wave velocity and ovum transport velocity in the rat oviduct".
Carolina Oses, Maria Paz Hernandez, Carlos Milovic, C. Llados, Manuel Villalon

FEDERATION AMER SOC EXP BIOL, USA, 2013

"High Levels of Leptin Reduces Ciliary Activity and Affects Ovum Transport Velocity in the Rat Oviduct".
Carolina Oses, Maria Paz Hernandez, Carlos Milovic, C. Llados, Manuel Villalon

44th annual meeting of the Society for the Study of Reproduction, USA, 2011

Development - PCL Processing Modules

Source code of these modules is avalaible per request.
These modules are provided as development processing tools, prior to their inclusion as standard modules in the current PixInsight Core distribution.
-----------------------------------------------------------------------------------------------------------------

TGV Module (Development release)

Collection of convolution processes.
Downloads: Windows - Linux


Included Processes:

Name: TGVDenoise
Description: Powerfull noise reduction tool. Now it includes a more flexible engine, capable of handling images with different noise models.
Category: NoiseReduction, TotalGeneralizedVariation

Name: TGVInpaint
Description: An inpainting algorithm. A mask tells the process wich pixels to replace, using first a precondition algorithm, and then TGV-based diffusion. Useful for removing unwanted objects (bad columns, blooming artifacts, stars, etc).
Category: Painting, TotalGeneralizedVariation

Name: TGVRestoration
Description: An deconvolution regularized with TGV. Gaussian and Poisson noise models implemented.
Category: Deconvolution, TotalGeneralizedVariation

-----------------------------------------------------------------------------------------------------------------

CMConv Module

Collection of convolution processes.
Downloads: Windows - Linux


Included Processes:

Name: Blur
Description: Very simple implementation of "average type" low pass filters. Now with an iterations parameter. Note the "Preserve luminance" option, to reduce chrominance noise.
Category: Convolution

Name: ButterworthFilter
Description: Band pass filter, that may work as a high pass (sharpen) or low pass (blur). Please note that this algorithm generates alliasing artefacts, specially the low pass version. Not recomended, but for educational purposes. Source code gives a nice starting point for Fourier analysis and filtering.
Category: Convolution

Name: GradientMapper
Description: This process calculates the gradient of a image (with user defined filters) and creates a new image with the gradient modulus. Gradient angle output is optional.
Category: Convolution

Name: Sharpen
Description: Simple implementation of laplacian like high pass filters. Good job with small scale contrast enhancement.
Category: Convolution

-----------------------------------------------------------------------------------------------------------------

CMGeneral Module

Collection of general purpose processes, mainly preprocessing.
Downloads: Windows - Linux


Included Processes:

Name: FFTRegister
Description: This is an implementation of the FFTRegistration script, by Juan C. with a wavelets optimization. In some cases, isolating a single wavelet layer generates best results, specially when noise is prominent at high frequencies. In the future this process will include rotations and scale variations too. Those algorithms are not working as expected, so only translations are implemented right now (which is still very useful in most multiple frames acquisitions).
Category: ImageRegistration

Name: SelectiveBlend
Description: Blend a grayscale image with color channels by user defined percentages. Useful for narrow band images blending.
Category: ColorSpaces

-----------------------------------------------------------------------------------------------------------------

CMIntensity Module

Collection of intensity transformations processes.
Downloads: Windows - Linux


Included Processes:

Name: GammaStretch
Description: A simple gamma stretch, with RTP.
Category: IntensityTransformations

Name: HistogramEqualization
Description: The goal of this process is to create a transfer curve that will generate a flat histogram. This means, pixel values will be distributed equally along the dynamic range. Fast and simple algorithm that works well in many daytime pictures (manually extract luminance). Not so useful with astronomical images. Future development will include histogram specification, that will allow better results in other applications.
Category: IntensityTranformations

Name: LogStretch
Description: Another function to stretch the pixel values, this time using logarithmic scales. Fast, and nice contrast.
Category: IntensityTranformations

Name: QuickSTF
Description: Instead of changing the STF by back and white points, and a MTF, this process defines two Brightness and Contrast parameters. An interactive mode lets the user to change the STF by clicking on the image, and the pixel coordinates will modify the SFT parameters. Funny image inspection.
Category: IntensityTranformations

-----------------------------------------------------------------------------------------------------------------

CMSegmentation Module

Collection of processes related to mask creation, object detection or feature extraction.
Downloads: Windows - Linux



Included Processes:

Name: CHT
Description: CHT stands for Circular Hough Transform. This algorithm transforms the image into a parametric space, where the intensity of the pixels is a measure of the probability of having a circle centered at that position. Different radius are calculated into different alpha channels.
Category: ObjectDetection

Name: ColorRange
Description: This process creates a new binary image that reflects the selection of a cube or spherical selection around a user defined "color". This selection may be defined in various space colors, giving a lot of freedom.
Category: MaskGeneration

Name: LHT
Description: CHT stands for Linear Hough Transform. This algorithm transforms the image into a parametric space, where the intensity of the pixels is a measure of the probability of having a straight line. X axis gives the inclination angle, and Y axis the absolute minimal distance from the origin (0.0).
Category: ObjectDetection

Name: ReadPSF
Description: Very simple implementation of the StarStatistics class, to evaluate PSF fitting by fitting gaussian functions to star shapes using a maximum likelihood approach.
Category: Image

Name: Seed
Description: More commonly known as the "Magic Wand". This algorithm plants a seed, and through local and global comparisons, determines a similar neighborhood.
Category: MaskGeneration

-----------------------------------------------------------------------------------------------------------------

HDR Module

Collection of processes related to high dynamic range compression.
Downloads: Windows - Linux


Included Processes:

Name: HDRCompresion (HDRC)
Description: Gradient Domain HDRC, using a multiscale gradient evaluation and attenuation function. Good job with daylight frames. Best results working with linear data. Astronomical usage needs refining, or creating a new attenuation algorithm based on this. Feedback is highly welcome.
Category: HDR

Name: HomomorphicFilter
Description: In the logarithmic scale, luminance and reflectance properties of the image may be separable by high ans low frequency components. A filtering process in the Fourier domain enhances local scale while comprising the dynamic range. Fast algorithm.
Category: HDR

-----------------------------------------------------------------------------------------------------------------

InterChannelCurves Module

Description: This process uses a user defined color space channel to define a amplification or attenuation factor, that will be applied to another user specified channel. The interface and internal algorithm is similar to the ColorSaturation process, but for any combination of channels.
Downloads: Windows - Linux


-----------------------------------------------------------------------------------------------------------------

ImportL Module

Description: Is there any process that changes the luminance of an image and generate chrominance artifacts? Or do you want to enhance saturation of zones after an HDR compression? Use this process to import the new luminance and modify the chrominance according to the changes. Intended for linear images, but non-linear images may benefit also using the special non-linearity parameter (roughly, it is the MTF used to stretch the image).
Downloads: Windows - Linux


-----------------------------------------------------------------------------------------------------------------

NotchFilter Module

Description: Similar to the DefectMap process, this tool operates in the fourier domain to remove noise in specific frequencies. The user must provide with frequency replacement map. Useful to remove sinusoidal patterns.
Downloads: Windows - Linux


-----------------------------------------------------------------------------------------------------------------

Spikes Module

Description: Just for fun, a simple low pass filter that emulates the physical phenomena that generates diffraction spikes in reflector scopes. Currently only creates spikes in the horizontal/vertical axis. It is recommended to use this in the linear stage. ***Alpha release***
Downloads: Windows - Linux


-----------------------------------------------------------------------------------------------------------------

TaylorRegister Module

Description: This is an experimental tool to register any kind of images, that present translations and/or rotations. The algorithm uses a Taylor series expansion to match images as if they were continuous functions. High SNR images are recommended.
Downloads: Linux