List of Projects

Dense Disparity Map Estimation (MSc Work)

Traffic monitoring (PhD Work)

Image-Sequence Registration

Vehicle Detection and Tracking

Post-Doc Work

Non-Rigid Registration of Stroboscopic Data (Image Series)

Microtubule Segmentation

CTF Simulator (GUI)

Defocus Estimation

Focal Series

Radiation Damage Studies

 

 


Presentation: Download

I have also made a presentation of this page. Presentation is in a PDF format. PDF contains embedded movies, which you need to have real player. I could not run the movies in Linux.



Dense Disparity Map Estimation (MSc Work)


 

Pollefeys

Fig (a)

 

Fig (b)

 

The objective in dense disparity map estimation is to obtain a textured metric 3D surface model starting with a sequence of free hand camera images (Figure a, taken from Pollefeys’ thesis). I investigated on how to find all corresponding pixels within epipolar image while dealing with high frequency regions, sparse structures and edge discontinuities (Figure b).

 

Publications

F. Samadzadegan, F. Karimi Nejadasl, Two Algorithms for Dense Disparity-Map Estimation of Close Range Images, Iranian Scientific Information Journal, 2003 (in Persian)

Center of Iran, Tehran, Iran, May 2003 (in Persian)

F. Samadzadegan, F. Karimi Nejadasl, Dense Disparity-Map Estimation of Close Range Images using Optical Flow, National Conference of National

Cartographic Center of Iran, Tehran, Iran, 2003 (in Persian)

F. Karimi Nejadasl, Automatic Dense Disparity-Map Estimation of Close-Range Images, Master Thesis, 2002 (in Persian)

 

 


Traffic monitoring (PhD Work)


 

Picture121.jpg

 

Picture003

image005

 

During my PhD, I worked with large image sequences recorded from a helicopter hovering above a freeway. The objective was to obtain the trajectories of each individual vehicle to model the behavior of the drivers. For this purpose, all images had to be registered and the camera motion had to be accounted for. Sequential image registration would suffer from error-propagation. A detailed modeling of the complete camera trajectory and the scene would be a complicated and computational expensive task. My research has shown that this problem can be tackled by making the model scene-depth independent using homography. I developed a framework for local registration of road areas in a robust and automatic way. As a part of the framework, the images were registered by searching in a parameter space. The search was done using differential evolution as a global optimizer.


An object identification method was developed to detect many moving vehicles with sparse details, different speeds, and low contrast with respect to their background. For each vehicle, robust tracking was achieved by a statistical analysis of all intermediate displacement results.

 

Image-Sequence Registration - Vehicle Detection and Tracking


 

image1.png

 

Movie 1 : lens corrected sequence

image1.png

                                                                                                   

Movie 2 : registered image-sequence of the road area

cutStab1.png

 

Movie 3 : cutted the road area

bf80.png

 

Movie 4: background / foreground identification

 

 

wcar1

 

Movie 5: Tracked white car

bcar1

 

Movie 6: Tracked dark car

 

Publications

F. Karimi Nejadasl, R.C. Lindenbergh, Robust Automatic Image-Sequence Stabilization of Road Area, manuscript under preparation

F. Karimi Nejadasl, R.C. Lindenbergh, Vehicle detection from an image sequence collected by a hovering helicopter, In Proceedings, WG III/5 Photogrammetric Image Analysis 2011, Germany

F. Karimi Nejadasl, A System for the Acquisition and Analysis of Image Sequences to Model Longitudinal Driving Behavior, PhD thesis, 2010

F. Karimi Nejadasl, B.G.H. Gorte, M. Snellen, S.P. Hoogendoorn, Evaluating Behavior of Objective Functions in Optimization based Image Sequence Stabilization in Presence of Moving Objects, In Proceedings of XXIth ISPRS Congress, Beijing, July 2008

F. Karimi Nejadasl, B.G.H. Gorte, S.P. Hoogendoorn, M.Snellen, Optimization Based Image Registration in Presence of Moving Objects, In Proceedings of Advanced School for Computing and Imaging, Heijn, June 2008

F. Karimi Nejadasl, B.G.H. Gorte, S.P. Hoogendoorn, M. Snellen, Optimization Based Image

Registration in Presence of Moving Objects, In Proceedings of International Calibration and Orientation Workshop, Castelldefels, February 2008

F. Karimi Nejadasl, B.G.H. Gorte, S.P. Hoogendoorn, Robust Vehicle Tracking in Video Images Being Taken from a Helicopter, In Proceedings of ISPRS, Commission VII, Enschede, May 2006

F. Karimi Nejadasl*, B.G.H. Gorte, S.P. Hoogendoorn, Optical Flow based Vehicle Tracking Strengthened by Statistical Decisions, ISPRS Journal of Photogrammetry and Remote Sensing, 61(3-4):159-169, 2006 (*corresponding author)

B.G.H. Gorte, F. Karimi Nejadasl, S.P. Hoogendoorn, Outline Extraction of a Motorway from Helicopter Image Sequence, In Proceedings: Stilla U, Rottensteiner F, Hinz S (Eds) CMRT05. IAPRS, XXXVI(Part 3/W24):179-184, August 29-30, 2005

 

 

 


Non-Rigid Registration of Stroboscopic Data (Image Series)


 

pts1_10_nap34_27Oct2010_Pos182

Superimposed the first image and the 10th image

Will be presented later

tnap34_27Oct2010_Pos182

Sum of the registered images (only translation)

nnap34_27Oct2010_Pos182

Sum of the registered images (non-rigid registration)

 

Future Publication

F. Karimi Nejadasl, M. Karuppasamy, A.J. Koster, R.B.G. Ravelli, Non-rigid image registration of Stroboscopic Cryo-Electron Microscopy Data, manuscript under preparation

 

 


Microtubule Segmentation


tomo1

 

Movie 1: tomogram

detectedMT1

 

Movie 2: detected Microtubules

 

cleanedPC

 

Convert detected MT to point cloud and clean the data

skeleton

 

Skeletonize point cloud (MT)

segmented

 

Segmented MT

 

 

 


CTF Simulator (GUI)


 

 

CTFExplorer.png

 

Code

Written in Matlab. Diplib library is required for 2D computation and optimization toolbox for first zero calculation.

 

 


Defocus Estimation


 

Fig2

(a) The first image and (b) the average of 10 images of a stroboscopic image series collected at defocus View the MathML source from the extracellular hemoglobin sample suspended in the carbon support hole.

 

 

Fig4.png

 

Defocus estimation for an image with carbon support. The requested defocus value was View the MathML source. (a) Log PS image, (b) FOM image, (c) Q-factor image. The minima are superimposed by yellow circles: (d) the black curve with dotted line represents the radial averaged log PS; CTF zeros are marked with vertical lines. The CTF zeros for the requested defocus value (View the MathML source) and for the estimated defocus are respectively shown in green (dashed) and red (continuous). The 1D log PS curve, depicted in dark blue, has been smoothed by a Gaussian filter (σ=7) for finding the minima. The same color-coding is used for the graphs of the radial averaged FOM (e) and Q-factor (f). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

 

Code

Written in matlab: required diplib, optimization and statistic toolbox

 

Publications

F. Karimi Nejadasl, M. Karuppasamy, A.J. Koster, R.B.G. Ravelli, Defocus Estimation from Stroboscopic Cryo-Electron Microscopy Data, Ultramicroscopy, 111(11): 1592-1598, 2011

 

 


Focal Series


registeredImg1

Movie 1: focal series

PSD1

Movie 2: Power spectrum of focal series

 

·        Simultaneously estimate defoci and amplitude contrast (Levenberg-Marquardt Algorithm)

·        Using all possible local minima of the radial and periodogram averaged power spectrum

·        Using linearity constraint between estimated defoci and requested defoci

·        Taking into account uncertainties by giving lower weight to unbailable defocus

 

background

Radial and periodogram average power spectrum of all focal series

Weiner

Wiener filtered image resulting from combination of all focal series with estimated defocus

 

 


Radiation Damage Studies


 

doseRate

Qualitative investigation of the dose-rate effect. The aligned and summed images of (a) and (e) low-flux, (b) and (f) medium-flux, (c) and (g) high-flux, and (d) and (h) high-flux short-exposure series are shown at two different integrated flux densities of (a)-(d) 50 e Å-2 and (e)-(h) 250 e Å-2, respectively. The scale bar shown in (a) corresponds to 30 nm

 

FOM

Plots of radial-averaged cosine phase error versus resolution for different dose rates. (a) Radial-averaged FOMs are given for a medium-flux series on Hb in a low-salt sample for integrated flux densities of 50, 100, 150, 200 and 250 e Å-2. (b) Close-up of (a) showing the first and second zero crossing of the CTF for a defocus of -3.37 µm. Radial-averaged FOMs for (c) the low-flux and (d) high-flux short-exposure series.

 

FSC

Fourier ring phase residual (FRPR) and Fourier ring correlation (FRC) as a function of dose. Medium-flux data were combined in groups of three images, corresponding to an integrated flux density of 15 e Å-2 per combined image. The first combined image was used as a reference.

 

 

Publications

M. Karuppasamy*, F. Karimi Nejadasl*, M. Vulovic, A.J. Koster, R.B.G. Ravelli, Radiation damage in single-particle cryo-electron microscopy: effects of dose and dose rate, Journal of Synchrotron Radiation, 18(3): 398-412, 2011 (*both authors contributed equally)