Showing papers by "Dimitris Maroulis published in 2015"

Grow-Cut Based Automatic cDNA Microarray Image Segmentation

Abstract: Complementary DNA (cDNA) microarray is a well-established tool for simultaneously studying the expression level of thousands of genes. Segmentation of microarray images is one of the main stages in a microarray experiment. However, it remains an arduous and challenging task due to the poor quality of images. Images suffer from noise, artifacts, and uneven background, while spots depicted on images can be poorly contrasted and deformed. In this paper, an original approach for the segmentation of cDNA microarray images is proposed. First, a preprocessing stage is applied in order to reduce the noise levels of the microarray image. Then, the grow-cut algorithm is applied separately to each spot location, employing an automated seed selection procedure, in order to locate the pixels belonging to spots. Application on datasets containing synthetic and real microarray images shows that the proposed algorithm performs better than other previously proposed methods. Moreover, in order to exploit the independence of the segmentation task for each separate spot location, both a multithreaded CPU and a graphics processing unit (GPU) implementation were evaluated.

A custom grow-cut based scheme for 2D-gel image segmentation

Abstract: This work introduces a novel method for the detection and segmentation of protein spots in 2D-gel images. A multi-thresholding approach is utilized for the detection of protein spots, while a custom grow-cult algorithm combined with region growing and morphological operators is used for the segmentation process. The experimental evaluation against four state-of-the-art 2D-gel image segmentation algorithms demonstrates the superiority of the proposed approach and indicates that it constitutes an advantageous and reliable solution for 2D-gel image analysis.

A GPU vs CPU performance evaluation of an experimental video compression algorithm

Abstract: Modern video compression algorithms put significant strain on a system's CPU, especially for video encoding. The ever increasing demands for using video compression algorithms in a wide range of applications necessitate the use of processing components that boost the speed and quality of the video compression algorithm's execution. The vast parallel computational capabilities of modern graphics processing units (GPUs) that usually remain underutilized makes them suitable for handling the processing load for video coding. This paper examines and evaluates the performance benefits of using the GPU over the CPU for an experimental video compression algorithm. An NVIDIA CUDA GPU implementation is evaluated against a traditional multithreaded CPU implementation. Experimental results show that at the highest resolution examined, the GPU approach achieved an impressive speedup ratio of 21.303x against the CPU for the decoding process, while for encoding, the speedup ratio reached 11.048x. Overall results indicate the prevalence of the GPU over the CPU, justified reasonably by the massive parallelism offered by the GPGPU computing paradigm, showing that the GPU should be the architecture of choice for high definition video coding.

Spot detection in 2d-gel electrophoresis images

Abstract: Introduction Two-dimensional gel electrophoresis (2D-GE) is a powerful and well-established method for separating complex protein mixtures according to their isoelectric points and their molecular weights [1]. The digital product of 2D-GE is grayscale images that may contain up to ~10.000 spots [2]. Detecting these spots is a challenging task of 2D-GE image analysis. The available software packages and techniques fail to detect some spots and detect a large number of spurious (falsepositive) spots. The proposed approach is effective in spot detection and outperforms the compared state-ofthe-art software packages. Method In order to detect spot center candidates, the proposed spot detection approach incorporates information based on the contourlet transform decomposition [3] and the regional intensity of 2D-GE images. Spot center candidates are then examined in order to eliminate multiple spot centers corresponding to the same spot. Results Several experiments utilizing real as well as synthetic 2D-gel image datasets containing a total of ~3000 spots were conducted in order to evaluate the performance of the proposed approach against two software packages (Delta2D [4] and Melanie 7 [5]). The results were statistically evaluated using sensitivity (S), precision (P), and the weighted harmonic mean of them, i.e. the Fmeasure (F). The proposed approach achieved an S value of 78%, outperforming Delta2D (74%), and Melanie 7 (69%). Regarding P, the proposed approach achieved 92%, compared to 77% for Delta2D and 95% for Melanie 7. For F, which is the most reliable measure, the proposed approach achieved a value of 84%, performing better than both Delta2D (75%) and Melanie 7 (80%). Detection results for a region of a real 2D-GE image for the proposed approach as well as for Delta2D and Melanie 7 are shown on Figure 1. It is evident that Delta2D and Melanie 7 missed some spots (yellow arrows) and also detected spurious spots (white arrows). On the other hand, the proposed approach detected all protein spots without detecting any spurious. Conclusions In this paper, an original approach for spot-detection in 2D-GE images is presented. The experimental results on real 2D-GE images demonstrate that it outperforms stateof-the-art software packages and is effective in detecting protein spots in 2D-GE images. (a) (b)

CVC: The Contourlet Video Compression algorithm for real-time applications

Abstract: Nowadays, real-time video communication over the internet through video conferencing applications has become an invaluable tool in everyone's professional and personal life. This trend underlines the need for video coding algorithms that provide acceptable quality on low bitrates and can support various resolutions inside the same stream in order to cope with limitations on computational resources and network bandwidth. In this work, a novel scalable video coding algorithm based on the contourlet transform is presented. The algorithm utilizes both lossy and lossless methods in order to achieve compression. One of its most notable features is that due to the transform utilised, it does not suffer from blocking artifacts that occur with many widely adopted compression algorithms. The proposed algorithm takes advantage of the vast computational capabilities of modern GPUs, in order to achieve real-time performance and provide satisfactory encoding and decoding times at relatively low cost, making it suitable for applications like video conferencing. Experiments show that the proposed algorithm performs satisfactorily in terms of compression ratio and speed, while it outperforms standard methods in terms of perceptual quality on lower bitrates.

A novel approach for accurate 2D-gel image segmentation

Abstract: This work introduces a novel 2D-gel image spot detection and segmentation method based on multithresholding, a custom grow-cut algorithm, region growing and morphological operators. Experimental results on real and synthetic data show that the proposed method outperforms state-of-the-art software solutions and results in more accurate segmentation.