Current work
- Fluorescence Microscopy (2006-2009)
- The confocal laser scanning microscope (CLSM) is an optical fluorescence microscope that scans sections of a specimen in 3D and uses a pinhole to reject most out-of-focus light. However, the quality of confocal microscope images suffers from two basic physical limitations. First, out-of-focus blurs due to the diffraction-limited nature of the optical microscope and secondly, the confocal pinhole drastically reduces the amount of photons detected by the photomultiplier causing poisson noise. The images produces by the CLSM can therefore benefit from post processing by deconvolution methods designed to reduce the blur and noise. Our current goal is to develop a fast and efficient algorithm for simultaneous estimation of the point-spread function (PSF) of the microscope and the specimen function. We realize that a good estimation of the PSF is very important for restoration of the original specimen. However, estimation of the PSF is an undetermined problem with no known unique solution. We overcome this problem by using a physical acquisition model of the microscope and by introducing a priori knowledge about the specimen. This stabilizes the estimation procedure and helps in deciding between candidate solutions.
- More info on P2R project website
Past work
- Nonstationary Signal Detection and Identification (2002-2005)
At the power system automation lab (PSAL) of Texas A&M University, I worked as a research assistant in collaboration with Mr. Karthick Muthu Manivannan (Research associate), Mr. Carl Benner (TEES research engineer), Dr. Peng Xu (postdoctoral researcher), and was advised by Dr. Don Russell (Regents and J. W. Runyon Professor) on the Electrical Power Research Institute (EPRI) project on Distribution Fault Anticipator. A brief history of this project is described in the Texas A&M Engineer Research Magazine and the abstract in the EPRI website.- The aim of this project was to develop an automatic recognition and classification algorithm for the data obtained from a power system distribution feeder monitoring unit. A state space approach was used to model the captured voltage and current waveforms, and the Kalman Filter and Wavelets to segment any disturbance which is present in the signal. This work is based on the premise that the morphology of the captured voltage or current signal's Root-Mean-Square (RMS) should also be used for automatically identifying and isolating the disturbances. The segmentation process divides the quasi-stationary RMS signal into pre-disturbance, disturbance and post-disturbance regions. The pre- and post-disturbance segments are essentially stationary while the nonstationarity nature is extracted as the disturbance segment. Each of these regions in the RMS of the signal can now be represented as strings or sequences of predefined wave patterns, called primitives. Any syntactically correct combination of these primitives will define the morphology of the mother RMS signal. The grammar and the model for each class is built from a set of positive examples or the learning set using the error-correcting grammatical inference (ECGI) learning technique. A stochastic extension of the ECGI similar to a Viterbi-like dynamic algorithm in combination with the k-nearest neighbors (kNN) algorithm is used to recognize the test captured data. The structure of a new pattern is learned by updating the current grammar with the inferred production rules.
- Archived website: past research work page
- Image Restoration (2001-2002)
- Non-linear filter like the median filter (MF) is useful for reducing random noise and periodical patterns, but direct median filtering have undesirable side effects such as smoothening of noise free regions, which results in loss of image detail and distortion of signal. Impulse noise is suppressed by selectively filtering the contaminated signal regions only, thus minimizing distortion of clean passages and loss of high frequencies. In the first phase, Support Vector Machines (SVM) are used to identify the set of pixels N that are likely to be contaminated by the mixed impulses. In the second phase, the image is restored by employing a combination of the best neighborhood match filter (BNM) and the multi-shell median filter (MMF) to these segmented regions. This method combines the effectiveness of the Best Neighborhood Matching (BNM) filter in suppression of the noise components while adapting itself to the local image structures, and the edge edge and finer image detail preserving characteristics of the MMF. To support our proposed method, numerical results are also provided, which indicate that the filter is extremely useful for preserving edges or monotonic changes in trend, while eliminating impulses of short duration.