Florin Rotaru

Bio: Florin Rotaru is an academic researcher from Romanian Academy. The author has contributed to research in topics: Optimization problem & Optic disc. The author has an hindex of 7, co-authored 42 publications receiving 150 citations.

Optic Disc Localization in Retinal Images

Abstract: The paper proposes an optic disc localisation method in color retinal images. It is a first step of a retinal image analysis project which will be completed later with other tasks as fovea detection and measurement of retinal vessels. The final goal is to detect in early stages signs of ophthalmic pathologies as diabetic retinopathy or glaucoma, by successive analysis of ophthalmoscopy images. The proposed method first detects in the green component of RGB image the optic disc area and then on the segmented area extracts the optic disc edges and obtains a circular optic disc boundary approximation by a Hough transform.

Parallel Processing and Bio-inspired Computing for Biomedical Image Registration.

Abstract: Image Registration (IR) is an optimization problem computing optimal parameters of a geometric transform used to overlay one or more source images to a given model by maximizing a similarity measure In this paper the use of bio-inspired optimization algorithms in image registration is analyzed Results obtained by means of three different algorithms are compared: Bacterial Foraging Optimization Algorithm (BFOA), Genetic Algorithm (GA) and Clonal Selection Algorithm (CSA) Depending on the images type, the registration may be: area based, which is slow but more precise, and features based, which is faster In this paper a feature based approach based on the Scale Invariant Feature Transform (SIFT) is proposed Finally, results obtained using sequential and parallel implementations on multi-core systems for area based and features based image registration are compared

Automatic multi-threshold image segmentation using metaheuristic algorithms

Abstract: In this paper is presented an automatic segmentation approach for gray level images based on usage of metaheuristic swarming algorithms for multiple thresholds computing The multi-threshold segmentation is an optimization problem while the thresholds must be determined and applied to the source image by minimizing an error measure Because the number of possible solution may be very large in case of multiple thresholds, we used four metaheuristic swarming algorithms to obtain faster the optimal solution of the segmentation problem: Bacterial Foraging, Particle Swarming, Multi Swarm and Firefly optimization As optimization criteria, root mean square error, peak signal-to-noise ratio and structural similarity index are used Each optimization algorithm allows obtaining the optimal solution in a reasonable number of iterations and the obtained results were compared

Inertial Data Based Learning Methods for Person Authentication

Abstract: A user authentication method based on biometric movement patterns is proposed. As input, the data acquired from the accelerometer of a wearable device is used. Using machine learning methods, i.e. k-Nearest Neighbors, Random Forests classifiers and a ID Convolutional Neural Network, the person which performs a known activity is identified from a set of 15 persons. Most approaches in the domain propose either the human activity type recognition or person identification using the gait pattern. In this research, the considered activities are: Walking and Working at computer. As features, sequences of 52 consecutive accelerations on the three axes are selected using the overlapping time window method. Even if computer work is a mainly static activity, the results obtained for identifying the person performing it are encouraging. The main advantage of the proposed method is that it does not involve computing of other features than those from the acceleration data.

Fireworks Algorithm Based Image Registration

Abstract: In the Image Processing (IP) domain, optimization algorithms have to be applied in many cases. Nature-inspired heuristics allow obtaining near optimal solutions using lower computing resources. In this paper the Fireworks Algorithm (FWA) behavior is studied for Image Registration (IR) problems. The IR results accuracy is analyzed for different types of images, mainly in case of pixel based registration using the Normalized Mutual Information. FWA is compared to Particle Swarming (PSO), Cuckoo Search (CSA) and Genetic Algorithms (GA) in terms of results accuracy and number of objective function evaluations required to obtain the optimal geometric transform parameters. Because the pixel based IR may fail in case of images containing graphic drawings, a features based IR approach is proposed for this class of images. Comparing to other nature inspired algorithms, FWA performances are close to those of PSO and CSA in terms of accuracy. Considering the required computing time, that is determined by the number of cost function evaluations, FWA is little slower than PSO and much faster than CSA and GA.

Fusion of Multispectral and Panchromatic Images Based on Morphological Operators

Abstract: Nonlinear decomposition schemes constitute an alternative to classical approaches for facing the problem of data fusion. In this paper, we discuss the application of this methodology to a popular remote sensing application called pansharpening, which consists in the fusion of a low resolution multispectral image and a high-resolution panchromatic image. We design a complete pansharpening scheme based on the use of morphological half gradient operators and demonstrate the suitability of this algorithm through the comparison with the state-of-the-art approaches. Four data sets acquired by the Pleiades, Worldview-2, Ikonos, and Geoeye-1 satellites are employed for the performance assessment, testifying the effectiveness of the proposed approach in producing top-class images with a setting independent of the specific sensor.

A Multi-threshold Segmentation Approach Based on Artificial Bee Colony Optimization

Abstract: This paper explores the use of the Artificial Bee Colony (ABC) algorithm to compute threshold selection for image segmentation. ABC is a heuristic algorithm motivated by the intelligent behavior of honey-bees which has been successfully employed to solve complex optimization problems. In this approach, an image 1D histogram is approximated through a Gaussian mixture model whose parameters are calculated by the ABC algorithm. For the approximation scheme, each Gaussian function represents a pixel class and therefore a threshold. Unlike the Expectation Maximization (EM) algorithm, the ABC based method shows fast convergence and low sensitivity to initial conditions. Remarkably, it also improves complex time consuming computations commonly required by gradient-based methods. Experimental results demonstrate the algorithms ability to perform automatic multi threshold selection yet showing interesting advantages by comparison to other well known algorithms.

Automatic Optic Disc Abnormality Detection in Fundus Images: A Deep Learning Approach

Abstract: Optic disc (OD) is a key structure in retinal images It serves as an indicator to detect various diseases such as glaucoma and changes related to new vessel formation on the OD in diabetic retinopathy (DR) or retinal vein occlusion OD is also essential to locate structures such as the macula and the main vascular arcade Most existing methods for OD localization are rule-based, either exploiting the OD appearance proper- ties or the spatial relationship between the OD and the main vascular arcade The detection of OD abnormalities has been performed through the detection of lesions such as hemorrhaeges or through measuring cup to disc ratio Thus these methods result in complex and in exible im- age analysis algorithms limiting their applicability to large image sets obtained either in epidemiological studies or in screening for retinal or optic nerve diseases In this paper, we propose an end-to-end supervised model for OD abnormality detection The most informative features of the OD are learned directly from retinal images and are adapted to the dataset at hand Our experimental results validated the effectiveness of this current approach and showed its potential application

A Comprehensive Review of the Fireworks Algorithm

Abstract: The fireworks algorithm, which is inspired from the phenomenon of fireworks explosion, is a special kind of swarm intelligence algorithm proposed in 2010. Since then, it has been attracting more and more research interest and has been widely employed in many real-world problems due to its unique search manner and high efficiency. In this article, we present a comprehensive review of its advances and applications. We begin with an introduction to the original fireworks algorithm. Then we review its algorithmic research work for single objective and multi-objective optimization problems. After that, we present the theoretical analyses of the fireworks algorithm. Finally, we give a brief overview of its applications and implementations. Hopefully, this article could provide a useful road map for researchers and practitioners who are interested in this algorithm and inspire new ideas for its further development.