Ciência da Computação
URI permanente desta comunidadehttps://repositorio.fei.edu.br/handle/FEI/342
Navegar
4 resultados
Resultados da Pesquisa
- A Bio-Inspired Methodology for Digital Imaging Forensic Detection(2019-09-05) SANTOS, G. L.; OLIVEIRA, G. R.; PRADO, F. F.; SERIKAKU, R.; SANTOS, R. M.; LOPES, G.; RODRIGUES, P. S.© 2019 IEEE.The increasing number of digital media users, as well as the development of multimedia platforms, such as smartphones and tablets, has also increased the number of users who professionally and fraudulently manipulate all types of digital media. This paper proposes a methodology based on an image processing pipeline to detect mainly copy-move type frauds, which are intended to hide or enlarge visual information. Our proposal has been tested in a Copy-Move Forgery database and the results were equal or better in performance than the state-of-art methods.
- Image Stitching Using Non-Extensive Statistics(2019-09-05) CARDOSO, E.; RISCH, H. A.; LAHERAS, L. P.; LUIZ, V.; RODRIGUES, P. S.; Guilherme Lopes© 2019 IEEE.Nowadays there are different ways to make image stitching with help of Fiducial Point Descriptors (FPD), whose find the matches between images for an application, such as SIFT and calculates the homography with RANSAC. However, by finding the right match when we have images on differents points of view could be difficult. This paper introduces the application of q-SFT, a newest variation of SFT in a stitch algorithm, that can recognize large viewpoint changes called as LVC.
- Recent Nature-Inspired Algorithms for Medical Image Segmentation Based on Tsallis Statistics(2020-03-09) LOPES, GUILHERME ALBERTO WACHS; SANTOS, R. M.; SAITO, N.T.; RODRIGUES, P. S.Recently, many algorithms have emerged inspired by the biological behavior of animal life to deal with complicated applications such as combinatorial optimization. One of the most critical discussions involving these algorithms is concerning their objective functions. Also, recently, many works have demonstrated the efficiency of Tsallis non-extensive statistics in several applications. However, this formalism has not yet been tested in most recent bio-inspired algorithms as an evaluation function. Thus, this paper presents a study of seven of the most promising bio-inspired algorithms recently proposed (a maximum one decade), from this entropy applied to the multi-thresholding segmentation of medical im- ages. The results show the range of values of q , the so-called non-extensivity parameter of the Tsallis entropy, for which the algorithms tested here have their best performance. It is also demonstrated that the Firefly algorithm (FFA) is the one that obtained the best per- formance in terms of segmentation, and Grey Wolf Optimizer (GWO) presents the fastest convergence.
- CARES 2.0: Completely automated robust edge snapper for cimt measurement in 300 ultrasound images-a two stage paradigm(2011) MOLINARI, F.; ACHARYA, U. R.; ZENG, G.; MEIBURGER, K. M.; RODRIGUES, P. S.; SABA, L.; SURI, J. S.The carotid intima-media thickness (IMT) is a widely used marker associated to the risk of cardiovascular diseases and to atherosclerosis progression. IMT measurement requires high accuracy and reproducibility. Computer-aided measurements improve accuracy and precision, but usually require user interaction. In this paper we proposed an improved method (called CARES 2.0) over the previously developed technique (called CARES 1.0). CARES 2.0 is a two stage process: Stage-I adapts an integrated approach of intelligent image feature extraction and line fitting for far adventitia border detection. Stage-II is a first order absolute moment (FOAM 1.0) coupled to a novel and improved heuristic search for the lumen-intima (LI) and media-adventitia (MA) peaks. CARES 2.0 brings in two novel scientific contributions: (a) ability to improve Stage-I to compare jugular vein versus carotid artery and (b) introduction bi-directional and robust FOAM. The improved method is a fully automated IMT measurement technique, and was validated on a multi-institutional database of 300 images exhibiting normal and pathologic carotids. We benchmarked CARES 2.0 against previously developed CALEX 1.0 and user-driven FOAM 1.0. CARES 2.0 showed an IMT measurement bias equal to -0.032±0.178 mm, which was better than CALEX 1.0 (0.070±0.331 mm), FOAM 1.0 (-0.091±0.161 mm) and CARES 1.0 (0.035±0.198 mm), respectively. Thus CARES 2.0 showed an improvement of 54% over CALEX 1.0, 65% over stand alone FOAM 1.0 and 9% over CARES 1.0. Compared to CARES 1.0, CARES 2.0 improved the reproducibility by 10%. CARES 2.0 ensured complete automation and increased the reproducibility of the IMT measurement, a step closer for clinical usage. Copyright © 2011 American Scientific Publishers.