filtro híbrido adaptativo en cascada para arreglos lineales … · nuevo filtro son...
TRANSCRIPT
Filtro Híbrido Adaptativo en Cascada para Arreglos Lineales Uniformes Orozco-Tupacyupanqui
Información Tecnológica Vol. 25 Nº 4 – 2014 103
Filtro Híbrido Adaptativo en Cascada para Arreglos Lineales Uniformes de Antenas Walter Orozco-Tupacyupanqui, Mariko Nakano-Miyatake y Héctor Pérez-Meana* Escuela Superior de Ingeniería Mecánica y Eléctrica, Instituto Politécnico Nacional, Av. Santa Ana 1000, Col. San Francisco Culhuacán, 04430 México D.F., México. ([email protected], [email protected], [email protected] ) * Autor a quien debe ser dirigida la correspondencia. Recibido Nov. 22, 2013; Aceptado Ene. 29, 2014; Versión final recibida Mar. 14, 2014 Resumen Este artículo propone un nuevo esquema de filtro híbrido adaptativo en cascada basado en los algoritmos de mínimos cuadrados promediados y recursivo de mínimos cuadrados para arreglos lineales uniformes de antenas para generar lóbulos de radiación. Los bloques adaptativos individuales usados para formar el nuevo filtro son interconectados a través del factor de arreglo lineal para simular un arreglo virtual de antenas y fijar la dirección deseada del lóbulo principal. La formulación completa del nuevo filtro se detalla y su capacidad de minimizar los efectos debidos a señales de interferencia y ruido blanco gaussiano aditivo es evaluada mediante simulación y comparada con otros filtros. Los resultados de la simulación mostraron que la rapidez de convergencia y las características de filtrado del nuevo modelo son mejores o similares a las de aquellos que lo constituyen. Se concluye que el nuevo esquema puede generar el lóbulo principal hacia la dirección de la señal deseada. Palabras clave: arreglo lineal de antenas, filtros adaptativos en cascada, filtros híbridos, lóbulo de radiación, minimización de interferencias
Adaptive Hybrid Filter in Cascade for Uniform Linear Antenna Arrays Abstract This paper proposes a new adaptive hybrid filter scheme based on the least mean square and the recursive least square algorithms for uniform linear antenna arrays to generate radiation beam patterns. The individual adaptive blocks used to form the new filter are interconnected by the linear array factor to simulate a virtual antenna array and to set the desired direction of the main beam pattern. The completed formulation of the new filter is detailed and its capability to minimize interference signals and adaptive white gaussian noise is evaluated by simulation and compared with other filters. The simulation results showed that the convergence speed and the filtering characteristics of the new model are better or similar to those that make it. In conclusion, the new scheme can generate the main beam pattern to the desired signal direction. Keywords: linear antenna arrays, cascade adaptive filters, hybrid filters, radiation beam pattern, interference minimization
Información Tecnológica Vol. 25(4), 103-122 (2014) doi: 10.4067/S0718-07642014000400013
Desarrollo de Algoritmos para Muestreo Compresivo Aplicado a Señales de Audio Moreno
Información Tecnológica – Vol. 25 Nº 5 2014 137
Desarrollo de Algoritmos para Muestreo Compresivo Aplicado a Señales de Audio Rodolfo G. Moreno(1),(2), Mauricio A. Martínez(1), Mariko Nakano(2), Héctor M. Pérez(2) (1) Universidad la Salle, Facultad de Ingeniería, Benjamín Franklin 47, C.P. 06140, México D.F.-México (2) Escuela Superior de Ingeniería Mecánica y Eléctrica, Instituto Politécnico Nacional. Av. Santa Ana 1000, C.P. 04430 México D.F. México (e-mail: [email protected]; [email protected]; [email protected]; [email protected]) Recibido Dic. 4, 2013; Aceptado Ene. 24, 2014; Versión final recibida Mar. 12, 2014 Resumen En este artículo se propone la aplicación de la transformada discreta de coseno (DCT) y el muestreo compresivo (MC) para la compresión de señales de audio. Usando el análisis espectral y las propiedades de la DCT se obtiene una señal de audio dispersa en el dominio de la frecuencia, lo cual permite el empleo de técnicas de MC para este tipo de señales en el dominio de la DCT. Así, se propone el uso de la DCT y el MC para obtener una representación eficiente de las señales de audio donde la DCT opera como un módulo de pre-procesamiento que permite obtener una representación dispersa de la señal en el dominio de la frecuencia, permitiendo la posterior aplicación del MC a las señales de audio. Con esto es posible obtener una representación de las señales de audio con menos muestras que las requeridas por el conocido teorema de muestreo. Palabras clave: muestreo compresivo, señales de audio, transformada discreta de coseno, dispersión en frecuencia, compresión
Development of compressive sampling algorithms applied to audio signals Abstract In this paper, it is proposed to apply the discrete cosine transform (DCT) and the compressive sampling (CS) techniques to audio signal compression. Using spectral analysis and the properties of the DCT, audio signals can be treated as sparse signals in the frequency domain. This is especially true for sounds representing tones. Thus, it is proposed using the DCT as a preprocessor to obtain a sparse representation in the frequency domain, combined with CS to obtain an efficient representation of the signals. It is shown that the subsequent application of CS represents the signals with less information than the well-known sampling theorem. Keywords: audio signals, compressive sampling, discrete cosine transform, frequency sparse, compression
Información Tecnológica Vol. 25(5), 137-146 (2014) doi: 10.4067/S0718-07642014000500019
Sistema de Reconocimiento de Rostros usando Visión Estéreo García-Rios
Información Tecnológica – Vol. 25 Nº 6 2014 117
Sistema de Reconocimiento de Rostros usando Visión Estéreo Elizabeth García-Rios, Enrique Escamilla-Hernández, Mariko Nakano-Miyatake y Héctor Pérez-Meana Escuela Superior de Ingeniería Mecánica y Eléctrica, Instituto Politécnico Nacional. Av. Santa Ana 1000, Col. San Francisco Culhuacán, 04430 México D.F. México (e-mail: ([email protected], [email protected], [email protected], [email protected]) Recibido Dic. 4, 2013; Aceptado Feb. 5, 2014; Versión final recibida Mar. 26, 2014
Resumen Este artículo propone un sistema biométrico usando imágenes estéreo que permite incrementar la seguridad de los sistemas de reconocimiento de rostros convencionales, al poder discriminar entre imágenes de rostro reales en 3D y falsificadas en 2D, empleando la información proporcionada por las imágenes estéreo. Una vez que el sistema determina que la imagen de entrada es un rostro real en 3D, cada imagen estéreo es caracterizada e independientemente procesada usando un método de extracción de características (EC) convencional. Seguidamente se emplea un método de fusión (MF) para combinar la información obtenida de ambas imágenes. Con el fin de determinar la combinación EC-MF más adecuada, se analizaron tres métodos de extracción de características y tres métodos de fusión. Los resultados experimentales obtenidos muestran que realizando la fusión de los datos extraídos mediante la transformación ondoleta discreta, antes de pasar a la etapa de clasificación, se obtienen los mejores resultados. Palabras clave: reconocimiento de rostros, visión estéreo, niveles de fusión, PCA, transformada ondoleta discreta, funciones de Gabor
Face Recognition system using Stereo Vision Abstract This paper proposes a biometric system using stereo images to improve the security performance of conventional face recognition systems, by discriminating between 3D real and forged 2D faces images, using the tridimensional information provided by the stereo images. Once the system determines that the face under analysis is an authentic 3D image, each stereo image is characterized and independently processed using three feature extraction methods. Then a fusion method is required to combine the information of both images. To this end three fusion methods are analyzed to achieve a more robust face recognition system. The experimental results show that by performing the fusion of the data extracted by wavelet transform, before going to the classification step, better results are obtained. Keywords: face recognition, stereo vision, fusion levels, PCA, wavelet transform, Gabor functions
Información Tecnológica Vol. 25(6), 117-130 (2014) doi: 10.4067/S0718-07642014000600015
Journal of ElEctromagnEtic WavEs and applications, 2016vol. 30, no. 13, 1712–1730http://dx.doi.org/10.1080/09205071.2015.1133328
Adaptive beamformer based on the augmented complex least mean square algorithm
Walter Orozco-Tupacyupanqui, Hector Perez-Meana and Mariko Nakano-Miyatake
postgraduate section-mechanical Electrical Engineering school, national polytechnic institute, mexico, mexico
1. Introduction
Complex-valued adaptive algorithms are frequently used in smart antenna arrays as a way to minimize the effects of distortion caused by interference signals and noise. For this reason, some adaptive algorithms such as the complex least mean square (CLMS) and the com-plex recursive least square (CRLS) have been developed to achieve this purpose.[1–4] These standard algorithms are able to simultaneously filter the real and imaginary components of the complex-valued signals involved in the adaptive array by using the covariance matrix R = E
[XXH
] of the input array vector X and the cross-covariance vector r = E
[d∗X
] between
the desired signal d and X . However, there is no guarantee that the second-order statistics of the signals can be fully captured by the least mean square (LMS) and recursive least square (RLS) adaptive beamforming systems. Moreover, in environments where the signals have been contaminated by White Gaussian noise and the desired signal is real-valued, the standard algorithms do not show an outstanding performance to reduce the effects of contamination in the imaginary component at the output of the adaptive array,[5] which in fact must be zero.
ABSTRACTIn this article, an adaptive beamforming system based on the augmented complex least mean square algorithm is analysed. In this approach, the adaptive filter is used as a widely linear system. The second-order statistical information of the signals involved in the array is exploited. Under this consideration, the ability of the adaptive array to minimize the effects of interferences and complex white noise could be enhanced. The equations for the optimal weights and the array factor are derived for the proposed beamforming system. Computer simulations have been performed to evaluate the performance of the adaptive array, and the results were compared with two of the most common standard adaptive beamforming algorithms: the least mean square and recursive least square. The numerical simulations show that the proposed adaptive array has a better performance in time and spatial domain as compared to the classical beamforming systems.
© 2016 informa uK limited, trading as taylor & francis group
KEYWORDSaugmented complex lms beamforming; widely linear beamformer (WlB); smart antenna array; second-order signal processing; lms algorithm; rls algorithm; radio communication systems
ARTICLE HISTORYreceived 11 June 2015 accepted 10 december 2015
CONTACT Hector perez-meana [email protected], [email protected]
J Supercomput (2016) 72:1291–1311DOI 10.1007/s11227-016-1646-6
A GPU implementation of secret sharing scheme basedon cellular automata
Rogelio Adrian Hernandez-Becerril1 · Ariana Guadalupe Bucio-Ramirez2 ·Mariko Nakano-Miyatake1 · Hector Perez-Meana1 ·Marco Pedro Ramirez-Tachiquin1,2
Published online: 25 March 2016© Springer Science+Business Media New York 2016
Abstract Secret sharing (SS) schemes basedon cellular automata (CA) are consideredas secure encrypting algorithms, where several secret data can be shared among somepersons. Recently the SS schemes can be applied to solve real-world problems, suchas security in cloud computing. The principal obstacle of use of the SS scheme is itsconsiderably high computational cost; especially if a large amount of secret data mustbe encrypted and shared. In this work, we propose a parallel CA-based SS schemesuitable for any kinds of digital data in the graphic processing unit using computeunified device architecture technology. The uses of global memory and shared mem-ory are analyzed from computational effectiveness and security points of view. Theexperimental results show the proposed parallel implementation provides a speeduprate more than 18-fold compared with its sequential implementation. Also we showthe increase of the security level of the parallel implementation with respect to thesequential implementation.
Keywords GPU · CUDA · Secret sharing · Cellular automata · Encryption
1 Introduction
Recently, numerous attacks on the cloud computing environment have occurred, leak-ing sensitive information to the public site and causing serious security flaw. Actually,
B Mariko [email protected]
1 Postgraduate Section of Mechanical Engineering School, Instituto Politecnico Nacional, AvenidaSanta Ana 1000, Coyoacán, San Francisco Culhuacan, Mexico City 04430, Mexico
2 UPIITA, Instituto Politecnico Nacional, Av. Instituto Politecnico Nacional 2580,Gustavo A. Madero, Barrio La Laguna Ticoman, Mexico City 07340, Mexico
123
Journal of Applied Research and Technology 13 (2015) 435-442
1665-6423/All Rights Reserved © 2015 Universidad Nacional Autónoma de México, Centro de Ciencias Aplicadas y Desarrollo Tecnológico. This is an open access item distributed under the Creative Commons CC License BY-NC-ND 4.0.
www.jart.ccadet.unam.mx
Available online at www.sciencedirect.com
Journal of Applied Research and Technology
*Corresponding author.E-mail address: [email protected] (H. Pérez-Meana).
Abstract
In almost all watermarking-based document authentication systems, the documents are considered as binary images and then, the watermark is embedded using some image watermarking algorithm. However actually important documents are saved using document file formats, such as Portable Document Format (PDF) or Open Document Format (ODF), among others, because in general the file size is smaller compared with an image file, and also these are considered as more secure than other types of file format. However, the documents with these formats can also be maliciously modified for illegal purposes, making necessary the development of mechanisms that are able to detect such modifications. Considering the situations mentioned above, this paper proposes a document authentication scheme in which a watermark is directly embedded into the document file format as part of the document itself. The experimental results show a desirable performance of the proposed algorithm.All Rights Reserved © 2015 Universidad Nacional Autónoma de México, Centro de Ciencias Aplicadas y Desarrollo Tecnológico. This is an open access item distributed under the Creative Commons CC License BY-NC-ND 4.0.
Keywords: Digital watermarking; Document authentication; Document file format; Script format document
Original
Script format document authentication scheme based on watermarking techniques
M. González-Leea, M. Nakano-Miyatakeb, H. Pérez-Meanab,*, G. Sánchez-Pérezb
a Universidad Veracruzana, Facultad de Ingeniería en Electrónica y Comunicaciones, Poza Rica, Veracruz, Méxicob Instituto Politécnico Nacional, Mechanical and Electrical Engineering School, México D.F., México
1. Introduction
Nowadays almost all documents are generated in a digital form and stored using some file formats such as Portable Document Format (PDF) or Open Document Format (ODF), among others, because of the reduction of storage space and rapid access that these file formats provide. However, the digital documents can be easily modified by unauthorized persons resulting in an altered document with the same quality as the original one. Despite some file formats such as PDF includes some security mechanisms, these can be broken as described in section 2. These tampered documents can be used or distributed illegally causing economical and moral damages to the involved persons. This fact suggests the necessity to develop efficient document authentication systems.
Several schemes have been developed to authenticate digital documents which embed invisible watermark into them; most of these schemes consider the digital documents as binary images. For example, Yang and Kot (2004) proposed a document authen-tication scheme, in which an authentication code is embedded by changing the spaces size between consecutive words and char-acters (Yang & Kot, 2004). The main drawback of this scheme is its high computational complexity and vulnerability against
noise contamination. Huang et al. (2004) proposed an authenti-cation method for binary images including text documents, in which firstly the binary image is segmented in blocks and then some pixels in each block are rearranged in order to enforce a given relationship between the number of black and white pix-els in it. During the authentication process, this relationship is verified for each block in order to authenticate each block. If the determined relationship is satisfied, then the block is considered as authentic, otherwise the block is labelled as tampered. The principal disadvantage of this method is that a degradation intro-duced in the encoded binary image is noticeable. Wu and Liu (2004) proposed block-wise binary image authentication scheme, in which flippable pixels in each block are manipulated in order to embed a watermark bit in that block (Wu & Liu, 2004). Here the embedded watermark is imperceptible, because the flippable pixels can be flipped without causing any distortion in the binary image. However, in general, the watermark embed-ding payload is very low compared with the number of flippable pixels into the image. To improve the embedding payload, Gou and Wu (2007) introduced the concept of “super-pixels” and wet paper coding into the Wu and Liu’s scheme (Gou & Wu, 2007). The “super-pixels” form a set of individually non-flippable pix-els, which can be removed or added together without causing visual distortion. Also Wu and Liu (2004) reported that their authentication scheme is robust to printing and scanning opera-
JART-12 Gonzalez-Lee.indd 435 23/06/15 11:46
Journal of Communication and Computer 11 (2014) 543-551 doi: 10.17265/1548-7709/2014.07.003
Cropping and Noise Resilient Steganography Algorithm
Using Secret Image Sharing
O. Ulises Juarez-Sandoval, AtoanyFierro-Radilla, Angelina Espejel-Trujillo, Mariko Nakano-Miyatake and
Hector Perez-Meana PostgraduateSection, ESIMECulhuacan, NationalPolytechnicInstitute, Mexico, Av.Santa Ana #1000, 04430,DelCoyoacan
Abstract: This paper proposes an image steganography scheme, in which a secret image is hidden into a cover image using a SIS (secret image sharing) scheme. Taking advantage of the fault tolerance property of the (k, n)-threshold SIS, where using any k of n shares (k ≤ n), the secret data can be recovered without any ambiguity. In order to increase the security of the secret information to digital attacks, the proposed steganography algorithm becomes resilient to cropping and impulsive noise contamination using SIS scheme. Among many SIS schemes proposed until now, the Lin and Chan’s scheme is selected as the SIS, due to its lossless recovery capability of a large amount of secret data. Stego-image quality and hiding capacity depend on the prim number used in polynomial. The proposed scheme is evaluated from several points of view, such as imperceptibility of the stego-image respect to its original cover image using the PSNR, quality of the extracted secret image, robustness of hidden data to cropping operation, impulsive noise contamination and the combination of both attacks. The evaluation results show a high quality of the extracted secret image from the stego-image when it suffered more than 20% cropping and/or high density noise contamination.
Key words: Steganography, data hiding, secret image sharing, cropping resilient, lagrange interpolation.
1. Introduction
Digital steganography is a technique to establish a
secret communication between sender and receptor
parts, hiding secret information into the carrier or cover
file, such as digital image, audio and video. The main
goal of the steganography is the generation of stego
files with secret data, which are perceptually and
statistically indistinguishable, from natural digital files,
in order to avoid arising any suspicion through its
transmission. Until now, several steganographic
algorithms have been proposed in the literature [1-5],
among them, the image stenographic techniques, in
which secret information is hidden into a carrier or
cover image, could be classified into two categories:
the first category is frequency domain steganography
and the second one is spatial domain steganography.
Generally, the frequency domain steganography offers
Corresponding author: Hector Perez-Meana Ph.D.,
professor, research fields: image processing, data hiding. E-mail: [email protected].
robustness of the hidden data, as Ref. [6], where a
robust steganography algorithm to JPEG compression
is proposed. This scheme uses the DCT transform and
QIM algorithm to make robust the secret image to
compression attack, while spatial domain
steganography offers higher capacity of the data
hiding.
The LSB (least significant bits) steganography is one
of the most popular spatial domain steganography, in
which the LSBs of the cover image are replaced by the
secret bits sequence. The maximum hiding capacity of
the LSBs steganography is limited by the number of
pixels of the cover image and if some regions of the
image are cropped or contaminated with additive noise,
the secret data cannot be recovered completely, in other
words, the LBSs image steganography schemes does
not have robustness to the cropping and the noise
contamination.
To increase the hiding capacity and the
imperceptibility of the LSB steganography, BPCS
(bit-plane complexity segmentation) steganography
D DAVID PUBLISHING
Circuits Syst Signal Processhttps://doi.org/10.1007/s00034-018-0820-x
Sparse Technique for Images Corrupted by MixedGaussian-Impulsive Noise
A. Palacios-Enriquez1 · V. Ponomaryov1 ·R. Reyes-Reyes1 · S. Sadovnychiy2
Received: 13 December 2016 / Revised: 16 April 2018 / Accepted: 19 April 2018© Springer Science+Business Media, LLC, part of Springer Nature 2018
Abstract In this paper, a novel framework is presented for denoising images thathave been corrupted by a mixture of additive and impulsive noise. The proposedmethod consists of three main stages: impulsive noise suppression, additive noisesuppression and post-processing. In the first stage, a pixel that has been contaminatedby impulsive noise is detected and filtered. In the next stage, filtering is based on sparserepresentation and 3D-processing using discrete cosine transform. Finally, the post-processing stage increases the filtering quality by using a bilateral filter and an edgerestoration technique. Evaluation is performed using objective criteria (PSNR andSSIM) and subjective human visual perception to confirm the methods performancecompared with state-of-the-art techniques.
Keywords Image denoising ·Additive noise · Impulsive noise ·Mixed noise · Sparserepresentation · PSNR · SSIM
B A. [email protected]
1 Instituto Politecnico Nacional, ESIME Culhuacan, Mexico City, Mexico2 Instituto Mexicano del Petroleo, Lazaro Cardenas 152, Mexico City, Mexico
1Abstract— Super-Resolution (SR) has many applications in several issues of the image processing by obtaining High-Resolution (HR) images from Low-Resolution (LR) images. In this paper, a SR technique that can increase the resolution in images of different nature is proposed. Our approach in obtaining SR image, first uses Lanczos interpolation of initial LR image, then edge features are extracted via convolution of an image with two different filters; following, the most informative features are performed employing principal component analysis (PCA). In next step, preprocessed image presented in blocks is used, where for an each block its sparse representation is performed using LR dictionary and another HR dictionary. In final step, the SR blocks are reconstructed resulting in improved SR image. Experimental results demonstrate the effectiveness of our method in comparison to state-of-the art techniques in terms of objective criteria PSNR, MAE and SSIM values as well as in subjective visual performance. Additionally, the proposed technique significantly reduces computational time in SR reconstruction.
Keywords— Super-Resolution, Sparse Representation, Dictionary, PCA, Interpolation, Edges, Filters, K-SVD.
I. INTRODUCCIÓN A SÚPER-RESOLUCIÓN (SR) de imágenes siempre es un tema importante en el procesamiento de imágenes, ya
que permite recuperar imágenes de Alta-Resolución (HR, por sus siglas en inglés) a partir de su correspondiente imagen de Baja-Resolución (LR, por sus siglas en inglés). Este tipo de tecnología tiene múltiples aplicaciones en distintas áreas como en medicina, ingeniería, visión por computadora, producción y posproducción de videos, televisión de alta definición (HDTV, por sus siglas en inglés), entre muchas otras aplicaciones civiles, industriales y científicas, debido a la alta demanda de imágenes HR [1].
En la mayoría de los casos, no es posible obtener dichas imágenes, debido a que el dispositivo electrónico con el que se capturó, esta imagen, posee sensores que no tienen una buena calidad o simplemente la imagen disponible es LR. Sí consideramos que la imagen LR es una versión de menor resolución de la imagen HR, sólo que decimada y con un efecto de emborronamiento, entonces la imagen LR puede ser modelada como en (1):
η+= DBYZ , (1)
V. A. Ramos, Instituto Politécnico Nacional (IPN), Ciudad de México, México, [email protected]
V. Ponomaryov, Instituto Politécnico Nacional (IPN), Ciudad de México, México, [email protected]
Y. Shkvarko, Centro de Investigación y de Estudios Avanzados del Instituto Politécnico Nacional (CINVESTAV-IPN), Guadalajara, México, [email protected]
R. R. Reyes, Instituto Politécnico Nacional (IPN), Ciudad de México, México, [email protected]
donde Y es la imagen HR desconocida, Z es la imagen LR, B es un filtro de emborronamiento, D es el operador de decimación y η es un ruido aditivo que normalmente se aproxima con distribución gaussiana [2].
Para resolver este problema, se han utilizado a lo largo de los años, diferentes herramientas matemáticas como la interpolación, transformadas (Fourier, Wavelet, etc.) y recientemente la representación dispersa (sparse representation).
De los métodos basados en interpolación destacan tres, que se han denominado, métodos clásicos; éstos son: la interpolación del vecino más cercano (NNI, por sus siglas en inglés), la interpolación bilineal [3] y la interpolación bicúbica [4]. Sobresaliendo la interpolación bicúbica, la cual obtiene la información de los pixeles faltantes mediante una regresión, utilizando un polinomio cúbico; este método se ha mantenido como una buena opción, no obstante, que al aplicar la interpolación bicúbica resulta en una imagen que presenta cierto emborronamiento en zonas de bordes.
La interpolación Lanczos [5] es otra interpolación ampliamente utilizada en señales digitales, la cual tiene resultados significativamente mejores con respecto a la interpolación bicúbica, esta interpolación utiliza el núcleo Lanczos, el cual se define como sigue:
2 2sin( )sin( / )
( )0
a x x a si a x aL x x
otro caso
π ππ
− < <°= ®°
, (2)
Dada una señal de dos dimensiones, como es el caso de una imagen, la fórmula de interpolación utilizando el núcleo presentado en (2) se da por:
1 1( , ) ( , ) ( ) ( )
x a y a
i x a j y aS x y s i j L x i L y j
+ +« » « »¬ ¼ ¬ ¼
= − + = − +« » « »¬ ¼ ¬ ¼
= − −¦ ¦ , (3)
donde ( , )S x y es la señal interpolada y ( , )s i j es la señal con la que se cuenta.
Otro enfoque ampliamente utilizado en SR de imágenes es el uso de transformadas, dentro de las que destaca la Transformada Discreta Wavelet (DWT, por sus siglas en inglés) [6]-[8], la cual descompone a la imagen LR en un conjunto de sub-bandas que representan los coeficientes de bajas y altas frecuencias contenidas en la imagen; sin embargo, en este enfoque se considera que la imagen LR ha sido obtenida mediante la misma DWT y corresponde a la sub-banda LL de la imagen HR, lo cual en términos prácticos, raramente se presenta.
La propuesta que se presenta a continuación tiene diversas ventajas en comparación con otros métodos y técnicas del
L
Image Super-Resolution via Block Extraction and Sparse Representation
V. A. Ramos, V. Ponomaryov, Member, IEEE, Y. Shkvarko, Senior Member, IEEE and R. R. Reyes, Member, IEEE
IEEE LATIN AMERICA TRANSACTIONS, VOL. 15, NO. 10, OCTOBER 2017 1977
Multimed Tools ApplDOI 10.1007/s11042-017-4968-3
Image super-resolution via two coupled dictionariesand sparse representation
Valentin Alvarez-Ramos1 ·Volodymyr Ponomaryov1 ·Rogelio Reyes-Reyes1
Received: 19 December 2016 / Revised: 12 May 2017 / Accepted: 20 June 2017© Springer Science+Business Media, LLC 2017
Abstract In image processing, the super-resolution (SR) technique has played an importantrole to perform high-resolution (HR) images from the acquired low-resolution (LR) images.In this paper, a novel technique is proposed that can generate a SR image from a single LRinput image. Designed framework can be used in images of different kinds. To reconstruct aHR image, it is necessary to perform an intermediate step, which consists of an initial inter-polation; next, the features are extracted from this initial image via convolution operation.Then, the principal component analysis (PCA) is used to reduce information redundancyafter features extraction step. Non-overlapping blocks are extracted, and for each block, thesparse representation is performed, which it is later used to recover the HR image. Using thequality objective criteria and subjective visual perception, the proposed technique has beenevaluated demonstrating their competitive performance in comparison with state-of-the-artmethods.
Keywords Super-resolution · Sparse representation · Feature-extraction · Filters ·Quality criteria
1 Introduction
Super-Resolution (SR) is one of the most active fields in image processing; the mainpurpose of SR is to obtain a High-Resolution (HR) image from a Low-Resolution (LR)
! Valentin [email protected]
Volodymyr [email protected]
Rogelio [email protected]
1 SEPI, ESIME Culhuacan, Instituto Politecnico Nacional, Av. Santa Ana no. 1000, Col. SanFrancisco Culhuacan, Mexico City, Mexico
A sub-block-based eigenphases algorithm with optimum sub-block size
Gibran Benitez-Garcia, Jesus Olivares-Mercado, Gabriel Sanchez-Perez, Mariko Nakano-Miyatake,Hector Perez-Meana ⇑Mechanical and Electrical Engineering School, National Polytechnic Institute of Mexico, Av. Santa Ana 1000, 04430 Mexico D.F., Mexico
a r t i c l e i n f o
Article history:Received 2 February 2012Received in revised form 4 May 2012Accepted 31 August 2012Available online 3 October 2012
Keywords:Face recognitionEigenphasesSub-block processingPCASVMPartial occlusion
a b s t r a c t
Several algorithms have been proposed for constrained face recognition applications. Among them theeigenphases algorithm and some variations of it using sub-block processing, appears to be desirable alter-natives because they achieves high face recognition rate, under controlled conditions. However, their per-formance degrades when the face images under analysis present variations in the illumination conditionsas well as partial occlusions. To overcome these problems, this paper derives the optimal sub-block sizethat allows improving the performance of previously proposed eigenphases algorithms. Theoretical andcomputer evaluation results show that, using the optimal block size, the identification performance of theeigenphases algorithm significantly improves, in comparison with the conventional one, when the faceimage presents different illumination conditions and partial occlusions respectively. The optimal sub-block size also allows achieving a very low false acceptance and false rejection rates, simultaneously,when performing identity verification tasks, which is not possible to obtain using the conventionalapproach; as well as to improve the performance of other sub-block-based eigenphases methods whenrank tests are performed.
� 2012 Elsevier B.V. All rights reserved.
1. Introduction
Face recognition is one of the most widely used biometric rec-ognition methods because it is the usual way used for most peopleto recognize another person. Additionally the data acquisitionmethod for face recognition is non-intrusive because it consistsonly in taking a picture, which can be carried out with or withoutcooperation of the person to be recognized. These facts do the facerecognition one of the biometric methods with higher acceptanceamong the users [1–3].
The face recognition as well as other biometric systems can beused for either, identity verification and person identification,depending on the data set used during the training stage. That isthe reason why it is important to clarify the differences amongthese two tasks. In the first case, the system is asked to determineif the person is who he/she claims to be, while during the identifi-cation task, the system is asked to determine the person, among aset of persons whose face characteristics are stored in the database,that more closely resembles the image under analysis. Thus therecognition task encompasses both identification and verification[4,5].
Fig. 1 shows the block diagram of a general face recognitionsystem, in which during the training stage, firstly the face image
is obtained. Subsequently in the feature extraction stage a set ofrelevant and near invariant features of the face image are extractedto generate a template of the person under analysis. During therecognition operation, depending on the required task, slightly dif-ferent processes are performed. Thus if the system is required toperform a verification task, besides the face image of the personunder analysis, the identity of the user must be provided, in orderto compare the estimated template of the person under analysiswith the template of the person who he/she claims to be in orderto take a decision; while if the system is required to perform anidentification task, it is necessary to provide the face image ofthe person to be identified in order to estimate the template(live-template) of the person under analysis and compare it withall stored templates in the database, in order to determine the useridentity.
In any face recognition system, the feature extraction stageplays a fundamental role because it is responsible for estimatinga set of reduced and almost invariant parameters that allows elim-inating the influence of variations related to hair style, illuminationchanges and partial occlusion; while maximizing the difference be-tween faces of different people [6–10]. Thus the effectiveness of aface recognition system strongly depends on the performance ofthe algorithm used in the feature extraction stage. Because of thatduring the last two decades, several feature extraction algorithmshave been proposed that intend to meet the requirements of pro-viding a smaller intra-person variability and larger inter-personvariability. Because the effect of an efficient feature extraction
0950-7051/$ - see front matter � 2012 Elsevier B.V. All rights reserved.http://dx.doi.org/10.1016/j.knosys.2012.08.023
⇑ Corresponding author. Fax: +52 55 5656 2058.E-mail addresses: [email protected] (M. Nakano-Miyatake), hmpm@prodigy.
net.mx, [email protected] (H. Perez-Meana).
Knowledge-Based Systems 37 (2013) 415–426
Contents lists available at SciVerse ScienceDirect
Knowledge-Based Systems
journal homepage: www.elsevier .com/ locate /knosys
Knowledge-Based Systems 110 (2016) 1–14
Contents lists available at ScienceDirect
Knowle dge-Base d Systems
journal homepage: www.elsevier.com/locate/knosys
Facial expression recognition with automatic segmentation of face
regions using a fuzzy based classification approach
Andres Hernandez-Matamoros a , Andrea Bonarini b , Enrique Escamilla-Hernandez
a , Mariko Nakano-Miyatake
a , Hector Perez-Meana
a , 1 , ∗
a Instituto Politecnico Nacional, Av. Santa Ana 10 0 0 Mexico D. F., 04430, Mexico b Politecnico di Milano, Via Ponzio 34/5, 20133 Milano Italy
a r t i c l e i n f o
Article history:
Received 28 February 2016
Revised 27 May 2016
Accepted 6 July 2016
Available online 7 July 2016
Keywords:
Robust facial expression recognition
2D Gabor functions
Automatic ROI segmentation
PCA
Low complexity classifier
Horizontal projective integral
a b s t r a c t
This paper proposes a facial expression recognition algorithm that automatically detects the facial image
contained in a color picture and segments it in two regions of interest (ROI)—the forehead/eyes and the
mouth—which are then divided into non-overlapping N ×M blocks. Next, the average of the first element
of the cross correlation between 54 Gabor functions and each one of the N ×M blocks is estimated to
generate a matrix of dimension L ×NM , where L is the number of training images. This matrix is then
inserted into a principal component analysis (PCA) module for dimensionality reduction. Finally, the re-
sulting matrix is used to generate the feature vectors, which are inserted into the proposed low com-
plexity classifier based on clustering and fuzzy logic techniques. This classifier provides recognition rates
close to those provided by other high performance classifiers, but with far less computational complexity.
The experimental results show that proposed system achieves a recognition rate of about 97% when the
feature vector from only one ROI is used, and that the recognition rate increases to approximately 99%
when the feature vectors of both ROIs are used. This result means that the proposed method can achieve
an overall recognition rate of approximately 97% even when one of the two ROIs is totally occluded.
© 2016 Elsevier B.V. All rights reserved.
1. Introduction
Facial expressions are one of the most powerful ways that hu-
man beings notify others of their emotional states. Approximately
55% of the messages related to an individual’s feelings are deliv-
ered via facial expressions [1] . Facial expressions are generated by
contracting facial muscles that temporally deform facial compo-
nents such as eyebrows, lips, the nose or mouth, etc. These expres-
sions have been grouped into seven basic classes that are univer-
sal across human ethnicities: anger, disgust, fear, happiness, sad-
ness, surprise and neutrality [1] . Facial expression recognition has
received a great deal of attention during the last decade because it
is an important tool when automatic interactions between humans
and machines are critical, such as in developing hospital nurse
robot assistants [2] , automatic animation, and intelligent tutoring
systems, among others [1 , 2] .
As a result, several efficient facial emotion recognition (FER)
algorithms have been developed. In many cases these use ap-
proaches already widely adopted for facial recognition applications
∗ Corresponding author.
E-mail address: [email protected] (H. Perez-Meana). 1 http://www.posgrados.esimecu.ipn.mx
such as Gabor functions, discrete wavelet transforms, local binary
patterns (LBP), Weber Local Descriptors (WLD), discriminant sparse
local spline approaches, and key points based feature extraction
methods [2–10] . Ali et al. [2] proposed using empirical mode de-
composition, which is a form of nonlinear and non-stationary data
analysis for facial emotion classification. The major advantage of
this technique, which has previously been used to detect epilep-
tic seizures using EEG signals to diagnose Alzheimer’s disease, is
that the basis function can be directly derived from the signal it-
self using local data characteristics, providing a fully data-driven
approach. In this study, the face image was first preprocessed to
segment it from the background using a rectangle based on a face
model [2] scaled to a fixed size, with its intensity normalized. Next,
a Radon transform of the resulting image was estimated and its
successive projections decomposed using the empirical mode de-
composition (EMD) approach. Then, the resulting data were fed
into three different dimensionality reduction schemes: Principal
Component Analysis (PCA) [11] combined with linear discriminant
analysis (LDA), PCA with local Fisher discriminant analysis (LFDA)
and kernel LFDA (KLFDA) [2] . Finally, the feature vectors obtained
from these schemes were fed into a classifier stage to make the fi-
nal decision about which of the seven types of facial expression
was present in the image under analysis. This system achieved
http://dx.doi.org/10.1016/j.knosys.2016.07.011
0950-7051/© 2016 Elsevier B.V. All rights reserved.
Microprocessors and Microsystems 45 (2016) 216–225
Contents lists available at ScienceDirect
Microprocessors and Microsystems
journal homepage: www.elsevier.com/locate/micpro
A compact digital gamma-tone filter processor
Areli Rojo-Hernandez
a , Giovanny Sanchez-Rivera
a , ∗, Gerardo Avalos-Ochoa
a , Hector Perez-Meana
a , Leslie S. Smith
b
a Instituto Politecnico Nacional ESIME Culhuacan, Av. Santana N 10 0 0, Coyoacan, 04260 Distrito Federal, Mexico b Computing Science and Mathematics University of Stirling, Stirling FK9 4LA, Scotland
a r t i c l e i n f o
Article history:
Received 14 January 2016
Revised 23 March 2016
Accepted 17 May 2016
Available online 18 May 2016
Keywords:
Auditory models
Cochlear implant processor
Gamma-tone Filter
a b s t r a c t
Area consumption is one of the most important design constrains in the development of compact digital
systems. Several authors have proposed making compact Cochlear Implant processors using Gamma-tone
filter banks. These model aspects of the cochlea spectral filtering. A good area-efficient design of the
Gamma-tone Filter Bank could reduce the amount of circuitry allowing patients to wear these cochlear
implants more easily. In consequence, many authors have reduced the area by using the minimum num-
ber of registers when implementing this type of filter. However, critical paths limit their performance.
Here a compact Gamma-tone Filter processor, formulated using the impulse invariant transformation to-
gether with a normalization method, is presented. The normalization method in the model guarantees
the same precision for any filter order. In addition, area resources are kept low due to the implemen-
tation of a single Second Order Section (SOS) IIR stage for processing several SOS IIR stages and several
channels at different times. Results show that the combination of the properties of the model and the
implementation techniques generate a processor with high processing speed, expending less resources
than reported in the literature.
© 2016 Elsevier B.V. All rights reserved.
1. Introduction
The development of digital artificial cochlear chips has attracted
the interest of engineers for developing portable applications such
as pitch detection, speech recognition and audio source localiza-
tion on mobile devices, or for auditory prostheses [1] . These ap-
plications have used a model of the biological cochlea due to its
capabilities for processing audio signals including natural sounds
[2] . This cochlea functions as a transducer, converting the mechan-
ical vibrations from the middle ear into electrical signals (auditory
nerve spikes). These signals are sent to the human auditory sys-
tem which responds to the information contained in the speech
and audio signals.
Several studies show that the sound processing carried out by
the cochlea, can be modeled using the over-complete Gamma-tone
filter-bank, due to its resemblance to the human auditory sys-
tem [2,3] . In addition, recently proposed mathematical models, re-
ported in the literature, show that Gamma-tone filter banks de-
signed using the impulse invariant transformation allow digital im-
plementation of the analogue cochlea while employing reasonable
∗ Corresponding author.
E-mail address: [email protected] , [email protected] (G. Sanchez-
Rivera).
computation with negligible distortion [4] . The hardware imple-
mentation of the cochlea, whether analogue or digital, is called
an artificial cochlea chip or silicon cochlea , [5] . Because an efficient
Cochlea chip is very important in several fields, the development
of efficient cochlea chips has been an active research field. One
of the first analogue silicon cochleae was developed by Lyon and
Mead [4] , using analogue VLSI 3 μm technology. It is reported that
this cochlea chip, implemented using a cascade of 480 bi-quad fil-
ter sections, provides similar behavior than the human cochlea. A
silicon cochlea which provides a good approximation of the hu-
man cochlea was proposed by Mandal et al. [6] . One of the most
recent approaches was focused on building a bio-realistic analog
CMOS Cochlea with high tunability and ultra-steep roll-off. The
Chip response has high fidelity with respect to physiological ex-
periments on mammalian cochlea and is 0.9 mm
2 in area and con-
sumes 59 . 5 − 90 . 0 μW [7] .
Analogue implementations of artificial Cochlea chips, such as
the above, are potentially efficient in terms of processing speed
and area when compared with the digital implementations. How-
ever, the analogue approach is susceptible to other factors, such as
temperature, transistor mismatch and power supply noise [7] . To
solve these problems several digital implementations of cochlear
chips have been proposed aiming for efficient sound processors
with minimal area. One of the critical factors to be considered in
http://dx.doi.org/10.1016/j.micpro.2016.05.010
0141-9331/© 2016 Elsevier B.V. All rights reserved.
A cheating-prevention mechanism for hierarchicalsecret-image-sharing using robust watermarking
Angelina Espejel-Trujillo1 & Mariko Nakano-Miyatake1 & Jesus Olivares-Mercado1 &
Hector Perez-Meana1
Received: 1 November 2014 /Revised: 13 May 2015 /Accepted: 18 May 2015 /Published online: 10 June 2015# Springer Science+Business Media New York 2015
Abstract Secret image sharing (SIS) techniques allow visual secrets to be shared betweenmultiple people. These techniques require a predetermined access structure to be satisfied toreveal the secret. SIS schemes have the potential to increase security in several applicationsincluding telemedicine and image transfer in the cloud computing environment, providingcontrolled access to confidential images. To date, a significant number of SIS schemes withvarious properties and access structures have been proposed. Among them, hierarchical SIS(HSIS) is considered foremost since almost all organizations and associations manage theirsecret information in a hierarchical manner. However, the HSIS scheme tends to suffer fromsecurity flaws as the secret image can often be visually leaked, even when the access structureis not properly satisfied. To avoid this security flaw, we propose a cheating preventionmechanism by introducing a robust watermarking technique based on the QuantizationIndex Modulation-Dither Modulation (QIM-DM) algorithm in the discrete cosine transform(DCT) domain. Experimental results demonstrate the superior effectiveness of our proposedcheating prevention mechanism. Furthermore, our approach retains all of the desirable prop-erties of the HSIS scheme.
Keywords Secret sharing . Secret image sharing . Hierarchical secret image sharing . Cheatingprevention .Watermarking . Authentication
Multimed Tools Appl (2016) 75:7855–7873DOI 10.1007/s11042-015-2701-7
* Mariko [email protected]
Angelina [email protected]
Jesus [email protected]
Hector [email protected]
1 Postgraduate and Research Section, Mechanical Electrical Engineering School, Instituto PolitecnicoNacional, Av. Santa Ana no. 1000, Col. San Francisco Culhuacan, Del. Coyoacan, Mexico City,Mexico
Multimed Tools Appl (2016) 75:9073–9094DOI 10.1007/s11042-015-2863-3
A scalable summary generation method basedon cross-modal consensus clustering and OLAPcube modeling
Gabriel Sargent1 ·Karina R. Perez-Daniel2 ·Andrei Stoian1 ·Jenny Benois-Pineau3 ·Sofian Maabout3 ·Henri Nicolas3 ·Mariko Nakano Miyatake2 ·Jean Carrive4
Received: 1 December 2014 / Revised: 21 July 2015 / Accepted: 3 August 2015 /Published online: 1 September 2015© Springer Science+Business Media New York 2015
Abstract Video summarization has been a core problem to manage the growing amount ofcontent in multimedia databases. An efficient video summary should display an overview ofthe video content and most existing approaches fulfill this goal. However, such an overview
� Gabriel [email protected]
Karina R. [email protected]
Andrei [email protected]
Jenny [email protected]
Sofian [email protected]
Henri [email protected]
Mariko Nakano [email protected]
Jean [email protected]
1 Vertigo-CEDRIC, CNAM, 292 rue St. Martin, 75141 Paris Cedex 03, France
2 SEPI, ESIME Culhuacan National Polytechnic Institute IPN, Unidad Profesional Adolfo LopezMateos, Edificio 7, Gustavo A. Madero, Linda Vista, 07738 Mexico City, Mexico
3 LaBRI, University of Bordeaux, Domaine universitaire, 351 cours de la Liberation,33405 Talence, France
4 Institut National de l’Audiovisuel - INA Expert, 4 avenue de l’Europe,94366 Bry-sur-Marne, France
Face region authentication and recovery system basedon SPIHT and watermarking
Clara Cruz-Ramos & Mariko Nakano-Miyatake &
Hector Perez-Meana & Rogelio Reyes-Reyes &Luis Rosales-Roldan
Received: 11 October 2013 /Revised: 18 February 2014 /Accepted: 4 April 2014 /Published online: 27 April 2014# Springer Science+Business Media New York 2014
Abstract The face regions of digital pictures are some of the principal target of tampering togenerate a potential scandal, causing social and economic damages to involved persons. In thispaper, we propose a face region authentication and recovery system, in which the face regionsare automatically protected at the moment when the picture is taken by a digital camera. Whenthe original face of the picture is replaced by another face by malicious person, the system candetect the tampered face and recover the original one. The proposed system consists of twostages: the face region protection stage and the face region tamper detection and recoverystage. In both stages, the face detection module based on the Viola-Jones algorithm, faceregion encoding/decoding modules based on the Set Partitioning in Hierarchical Trees(SPIHT) algorithm and watermarking module based on Quantization Index Modulation(QIM) are used. These three algorithms, Viola-Jones detector, SPIHT and QIM, are determinedas most adequate algorithms for proposed system after several evaluations. The experimentalresults show a high quality of the watermarked as well as the recovered images, obtainingaverage Peak Signals to Noise Ratios (PSNR) of more than 40 and 38 dB, respectively.
Keywords Face detection . Tamper detection . Self-recovery .Watermarking . SPIHT. QIM
1 Introduction
Nowadays almost all newspapers offer their electronic versions in Internet, in which digitalimages have very important roll to provide news or articles. Also in social networks, manypeoples publish their digital images together with some comments, because generally visualinformation given by images has more impact than the written information. However, digitalimages can be modified easily using computational tools, such as Photoshop and Corel Draw,without any visual artifact. In many cases, tampered images cause economic and socialdamages to the involved persons in the images. Considering the above situation, the
Multimed Tools Appl (2015) 74:7685–7709DOI 10.1007/s11042-014-2006-2
C. Cruz-Ramos :M. Nakano-Miyatake (*) : H. Perez-Meana : R. Reyes-Reyes : L. Rosales-RoldanMechanical Electrical Engineering School, Instituto Politécnico Nacional, Av. Santa Ana no. 1000, Col. SanFrancisco Culhuacan, Mexico, DF, Mexicoe-mail: [email protected]