Back to the roots! Before obtaining PhD degree in the area of machine vision I designed automation systems. This year I "returned back to the roots" (for a shortwhile only) and designed a complete automation system controlling a MVR (mechanical vapour recompression) evaporator used for concentrating whey permeate. The PLC used controlling the evaporator is a Siemens S7 CPU 315-2 PN/DP with more than 200 inputs/outputs. OIU used for controlling the system is a WinCC v7.3.

Starting from 11th of August 2014 I have been working at Fuel-3D, a company which currently produces one of the most accurate 3D scanners in the market in its price category.

During the past few days I have virtualized different versions of Windows OSes using VMware player, the host being a Linux (Ubuntu distribution), and I need to change the physical network adapter in the bridged mode.

I just noticed that there is a bug in the code that manifests itself in "later" versions of Matlab. The bug is related to the Matlab function called IMFILTER. What happens is that when calculating the diffusion weights if the first parameter is in single float format (and not double) then the results are unstable. By unstable I mean that the actual numbers change when you apply the function on the same data. I suspect that it is related to the way how memory is reserved in the MEX-code.

Here it is finally, the segmentation code that I had promised quite long time ago! The Matlab code segments stereo disparity maps using a model based on implicit dynamic surfaces (also known as level sets). The code itself is a combination of Matlab and MEX (Matlab EXecutable), programmed in C, in order to improve the efficiency. The segmentation algorithm, and the related mathematics, are explained in my phd thesis. Therefore, in order to have a better understanding how it works I encourage you to have a look there as well.

Time is ripe to change the look and feel of my web-site so that it will be more accessible and easier to navigate! I have changed the URL-structure of the site, so some of the old links might not work properly temporily, although most of the old links have been permanently redirected with HTTP status code 301. Please bear with me, I will try to get the changes implemented as soon as possible!

Abstract: Correspondence techniques start from the assumption, based on the Lambertian reflection model, that the apparent brightness of a surface is independent of the observer's angle of view. From this, a grey value constancy assumption is derived, which states that a change in brightness of a particular image pixel is proportional to a change in its position. This constancy assumption can be extended directly for vector valued images, such as RGB. It is clear that the grey value constancy assumption does not hold for surfaces with a non-Lambertian behaviour and, therefore, the underlying image representation is crucial when using real image sequences under varying lighting conditions and noise from the imaging device. In order for the correspondence methods to produce good, temporally coherent results, properties such as robustness to noise, illumination invariance, and stability with respect to small geometrical deformations are all desired properties of the representation.

Abstract: Cell migration, formation of cellular protrusions (e.g. blebs, filopodia), and structural reorganization are important phenomena in cell biology. Precise quantifications of movement/deformation are crucial to understand these processe at different levels of organization. We apply computer vision methods for combined optical flow (OF) and muti‐scale (MS) motion estimation of membrane translations, end growing and protrusion formation in fluorescence microscopy images. For these cases we bound OF error and optimal sampling rate, in order to guide biologists on their experimental conditions. We also show the advantages of OF methods compared with manual segmentation and tracking.

I have made an example, with Matlab code included, related to both geometric- and geodesic level sets models. The example shows how the boundary separating inner- and outer parts of the segment can be driven temporally towards the image boundaries.

Here it is, finally! Interested in segmenting disparity maps, or perhaps about robust image representation spaces for disparity calculation, or about variational disparity or optical-flow calculation? All this information, including one complete chapter explaining how the related models can be solved effiiently, can be found in the complete version of my phd thesis!

Specialists Create a System that Analyses Male Infertility via Internet

Infertility affects roughly 15% of couples in Chile. It is estimated that annually around 6000 enquiries, related to infertility, are made to medical experts. However, not all the couples have the possibility of consulting an expert. Chile stretches over 4630km, is only 430km at the widest and has a population of 17 million. Therefore, the geography itself presents difficulties. In order to solve the problem, scientists at Instituto de Neurociencia Biomédica (Institute of Biomedical Neurosciences) have developed a centre of digital semen analysis based on Internet.

PUBLISHED!

Paper "FROM SENSORS TO SPIKES: EVOLVING RECEPTIVE FIELDS TO ENHANCE SENSORIMOTOR INFORMATION IN A ROBOT-ARM", by Dr. (soon to be) Niceto Luque Sola and Dr. Jesús Garrido, has been accepted for publication in the International Journal of Neural Systems.