Many existing researches on image watermarking for copyright protection need to use original image in retrieving watermark. Though it is more robust, it would cause some problems about the authorization of original image. In this thesis, we propose a method based on DCT domain without using original image. Using the property of concentrating energy in DCT transform, the energies of blocks are used for further processing. In the embedding algorithm, the DC coefficients of blocks are first collected. Then they are divided by some number to get remainders. The values of embedded data are embedded in the relationship between corresponding location of embedded data and other locations by adjusting the remainders in all locations. Some typical watermarking attacks and noise are used to evaluate the robustness of our method. Compared with other competing algorithms, it shows that the survival rate of watermark in our method can be almost the same or even better then those methods which need original image. The error rate of the lowest quality JPEG compression can be adjusted less then 1%, when the length of embedding data is 512 bits. Our proposed method can be further improved by adjusting the values of remainders and the block size. These provide flexibility to satisfy different requirements.

Submitted by Fabio Sobreira Campos da Costa (fabio.sobreira@ufpe.br) on 2016-06-21T15:14:55Z No. of bitstreams: 2 license_rdf: 1232 bytes, checksum: 66e71c371cc565284e70f40736c94386 (MD5) Vitor_de_Andrade_Coutinho-dissertacao_ppgee.pdf: 3622975 bytes, checksum: 01a22e0302dfc1890d745c6b1bffe327 (MD5) / Made available in DSpace on 2016-06-21T15:14:56Z (GMT). No. of bitstreams: 2 license_rdf: 1232 bytes, checksum: 66e71c371cc565284e70f40736c94386 (MD5) Vitor_de_Andrade_Coutinho-dissertacao_ppgee.pdf: 3622975 bytes, checksum: 01a22e0302dfc1890d745c6b1bffe327 (MD5) Previous issue date: 2015-02-23 / CNPq / O presente trabalho aborda o desenvolvimento de aproximações para a transformada dis- reta do osseno (DCT) utilizando a abordagem pruning. Devido à propriedade da ompa ta- ção de energia, a DCT é empregada em diversas apli ações de ompressão de dados. Embora algoritmos rápidos permitam omputar a DCT e ientemente, operações de multipli ação são inevitáveis. Devido a res ente demanda por métodos de baixo onsumo energéti o, novos algoritmos de usto omputa ional reduzido são ne essários. Neste ontexto, aproximações para a DCT foram propostas nos últimos anos. Tais aproximações permitem algoritmos livres de multipli ação, sem a ne essidade de operações de ponto utuante, mantendo o desempe- nho de ompressão omparável ao forne ido por métodos baseados na DCT. Uma abordagem adi ional para reduzir o usto omputa ional da DCT é a utilização de pruning. Tal té ni a onsiste em não onsiderar oe ientes dos vetores de entrada e/ou saída que apresentam menor relevân ia em termos de energia on entrada. No aso da DCT, esses oe ientes são os termos de mais alta frequên ia do vetor transformado. A apli ação de pruning a aproxima- ções para a DCT é uma área pou o explorada. O objetivo deste trabalho é apli ar a té ni a a diferentes métodos aproximados para a DCT. As transformações resultantes foram apli adas no ontexto de ompressão de imagem e vídeo e os resultados mostraram desempenho ompa- rável ao de métodos exatos a um usto omputa ional bastante reduzido. Uma generalização do on eito é apresentada, assim omo uma análise da omplexidade aritméti a. / This work introdu es approximate dis rete osine transforms (DCT) based on the pruning approa h. Due to the energy ompa tion property, the DCT is employed in several data ompression appli ations. Although fast algorithms allow an e ient DCT omputation, multipli ation operations are inevitable. Due to the in reasing demand for energy e ient methods, new algorithms with redu ed omputational ost are required. In this ontext, DCT approximations have been proposed re ently. Su h approximations allow multipli ation free algorithms whi h an avoid oating point operations, while maintaining a ompetitive performan e. A further approa h to redu e the omputational ost of the DCT is pruning. The te hnique onsists of dis arding input and/or output ve tors oe ients whi h are regarded as less signi ant. In the ase of the DCT, su h oe ients are output oe ients asso iated to higher frequen y terms. Pruned DCT approximations is a relatively unexplored eld of resear h. The obje tive of this work is the ombination of approximations and pruning to derive extremely low- omplexity DCT approximations. The resulting methods were applied in the image and vídeo ompression s enario and results showed omparative performan e with exa t methods at a mu h lower omputational omplexity. A qualitative and quantitative omparison with a omprehensive list of existing methods is presented. A generalization of the pruning on ept is presented.

Ce mémoire constitue l'aboutissement du projet de recherche de Patrick Keroulas et aborde la notion de compression vidéo, domaine en pleine ébullition avec la démocratisation de l'équipement vidéo et des réseaux de télécommunication. La question initiale est de savoir s'il est possible de modifier le contenu de l'image directement dans un flux binaire provenant d'une séquence vidéo compressée. Un tel dispositif permettrait d'ajouter des modifications en n'importe quel point d'un réseau en évitant le décodage et recodage du flux de données, ces deux processus étant très coûteux en termes de calcul. Brièvement présentés dans la première partie, plusieurs travaux ont déjà proposé une gamme assez large de méthodes de filtrage, de débruitage, de redimensionnement de l'image, etc. Toutes les publications rencontrées à ce sujet se concentrent sur la transposition des traitements de l'image du domaine spatial vers le domaine fréquentiel. Il a été convenu de centrer la problématique sur une application potentiellement exploitable dans le domaine de la télédiffusion. Il s'agit d'incruster un logo ajustable en position et en opacité dans un fichier vidéo codé avec la norme MPEG-2, encore couramment utilisée. La transformée appliquée par cet algorithme de compression est la DCT (Discrete Cosine Transform). Un article publié en 1995 traitant de la composition vidéo en général est plus détaillé car il sert de base à cette étude. Certains outils proposés qui reposent sur la linéarité et l'orthogonalité de la transformée seront repris dans le cadre de ce projet, mais la démarche proposée pour résoudre les problèmes temporels est différente. Ensuite, les éléments essentiels de la norme MPEG-2 sont présentés pour en comprendre les mécanismes et également pour exposer la structure d'un fichier codé car, en pratique, ce serait la seule donnée accessible. Le quatrième chapitre de l'étude présente la solution technique mise en oeuvre via un article soumis à IEEE Transactions on Broadcasting. C'est dans cette partie que toutes les subtilités liées au codage sont traitées : la structure en blocs de pixel, la prédiction spatiale, la compensation de mouvement au demi-pixel près, la nécessité ou non de la quantification inverse. À la vue des résultats satisfaisants, la discussion finale porte sur la limite du système : le compromis entre son efficacité, ses degrés de liberté et le degré de décodage du flux.

My research is concentrated on applications of linear algebra in engineering, including wavelet analysis and structured matrices and (currently) approximation of large dense matrices by a mosaic of low rank blocks. / Singapore-MIT Alliance (SMA)

There has been a great progress in the field of graphics processors. Since, there is no rise in the speed of the normal CPU processors; Designers are coming up with multi-core, parallel processors. Because of their popularity in parallel processing, GPUs are becoming more and more attractive for many applications. With the increasing demand in utilizing GPUs, there is a great need to develop operating systems that handle the GPU to full capacity. GPUs offer a very efficient environment for many image processing applications. This thesis explores the processing power of GPUs for digital image compression using Discrete cosine transform.

In embedded system, multimedia application is more important than before, And these products appearing more often. In addition, handholding devices are more and more popular, so these products¡¦ price is usually chip than others, and they concern more about power consumption. So in our design, we can¡¦t only focus on performance, low power and low cost are become the one of the most important factors. The main contribution of this thesis is that in MP3 multimedia application we analyzed ¡Bestimated and optimized our hardware and software to achieve low power and low cost issues. In software part, we used the skill of optimization to optimize our complied assembly codes. In hardware part, we analyzed the MP3 decode algorithm, found the critical part, and implement them with hardware to hope we can use the smallest hardware cost, to achieve highest acceleration. And we hope that through our research, we can establish a fundament about development a special application platform.

Wireless video systems have applications in cellular videophones, surveillance systems, and mobile patrols. The design of a wireless video system must consider two important constraints: low bit rate and low power dissipation. The ITU-T H.263 video codec standard is suitable for low bit rate wireless video systems, however it is computationally intensive. Some of the most computationally intensive operations in H.263 are the Discrete Cosine Transform (DCT) and the Inverse Discrete Cosine Transform (IDCT), which perform spatial compression and decompression of the data. In an ASIC implementation of H.263, the high computational complexity of the DCT and IDCT leads to high power dissipation of the blocks. Low power design of the DCT and IDCT is essential in a portable wireless video system. This paper examines low power design techniques for DCT and IDCT circuits applicable for low bit rate wireless video systems. Five low power techniques are applied to baseline reference DCT and IDCT circuits. The techniques include skipping low energy DCT input, skipping all-zero IDCT input, low precision constant multipliers, clock gating, and a low transition data path. Gate-level simulations characterize the effectiveness of each technique. The combination of all techniques reduces average power dissipation by 95% over the baseline reference DCT and IDCT blocks. / Master of Science

La compréhension des mécanismes de déformation dans les matériaux polycristallins est un problème important, qui conditionne notre capacité à concevoir et à produire des pièces de structure plus sures et avec un impact environnemental moindre. Cette compréhension est aujourd'hui limitée par notre capacité à observer à la fois la microstructure du matériau et ses mécanismes de déformation en trois dimensions (3D) aux petites échelles, et à informer les simulations mécaniques à partir des mécanismes physique de déformations du réseau cristallin. Des progrès considérables ont été faits dans les dernières décennies avec les observations de surfaces (i.e. technique EBSD associée a de la corrélation d’image) qui a permis de nombreuses études combinant des observations expérimentales à des simulations, à partir de la surface de la microstructure. Cependant, une comparaison précise sans connaitre la microstructure sous-jacente reste un défi. Dans ce travail, nous proposons une nouvelle méthodologie basée d'une part sur des mesures couplant la tomographie et la diffraction des rayons X, et d'autre part sur des simulations mécaniques de platicité cristalline. Cette approche permet une comparaison quantitative en volume entre les mécanismes de déformation, l’évolution de la courbure du réseau cristallin et les champs mécaniques simulés.Pour ce faire, une machine de traction dédiée aux expériences 4D d’imagerie par diffraction sur grands instruments a été conçue, et utilisée pour déformer en tension un échantillon d’Aluminium Lithium. La cartographie 3D de la microstructure a été obtenue par tomographie par contraste de diffraction, et un agrégat de trois grains dans le volume de l’échantillon a été choisi comme région d’intérêt pour des observation 4D par topotomographie. L’apparition des premières bandes de glissement en volume et leur évolution au cours du chargement ont été observées le long de plans cristallographiques bien définis. Les trois grains ont montré une activité plastique le long de deux familles de plans différents, pas toujours en accord avec une analyse macroscopique du facteur de Schmid, ce qui est attribué à l'influence du voisinage sur l'activation des systèmes de glissement. Les changements d’amplitude et d’orientation de la courbure moyenne des grains ont été mesures avec un niveau de détail sans précédent, par une analyse tridimensionnelle des courbes de reflexions.En parallèle, des simulations de la plasticité cristalline par éléments finis (CPFE) ont été menées utilisant la cartographie tridimensionnelle de la microstructure mesurée expérimentalement. Un chargement uniaxial de traction a été applique pour reproduire numériquement l’expérience, et comparer grain par grain l’activité plastique. L’activité des systèmes de glissement prédite par le modèle est conforme aux observations expérimentales d’une activité plastique le long de deux plans. Un cadre mathématique pour prédire l’angle de Bragg local en fonction des déformations et des rotations du réseau cristallin a été formulé. Un post-traitement des champs intragranulaires de déformation à partir des résultats des simulations CPFE a montré une excellente concordance avec les résultats expérimentaux. Ce résultat confirme que la topotomographie in-situ aux rayons X est un outil prometteur pour l’étude des premiers stades de la plasticité cristalline en volume. / Understanding the intimate details of plastic deformation in polycrystalline materials is an important issue to improve material design and ultimately produce safer structural parts with less impact on the environment. This understanding is presently limited by our ability to observe both the microstructure of the material and the deformation processes in three dimensions (3D) at small length scales and inform mechanical simulations with physical deformation mechanisms of the crystal lattice. Considerable progress has been made in the last decade with surface observation (eg EBSD coupled to digital image correlation) which led to numerous studies combining experimental observations and simulations from the surface microstructure. However, an accurate comparison without knowing the underlying microstructure remain challenging. In this work, we propose a new methodology which allows a quantitative comparison between the observation of deformation mechanisms, the evolution of the grain lattice curvature and the simulated mechanical fields.For that purpose, a mechanical stress rig dedicated for synchrotron 4D diffraction imaging experiments has been designed, and used to deform an Aluminium-Lithium specimen under tension. The 3D grain map has been obtained by diffraction contrast tomography analysis, and a cluster of three grains within the bulk has been selected to be the region of interest of the 4D observation by X-ray topotomography. The appearance and evolution of 3D crystalline defects as a function of the applied load has been observed to be located along well defined crystallographic planes. All three grains showed plastic activity on along two different set of planes, which is not always coherent with a macroscopic Schmid Factor analysis. The change of the amplitude and the orientation of the average grain curvature has been measured with an unprecedented level of detail by means of 3D rocking curve analysis.In parallel, crystal plasticity finite element (CPFE) simulations have been carried using the 3D grain map measured experimentally. Tension loading was applied to reproduce the experiment numerically and compare the plastic activity on a grain by grain basis. The slip system activity predicted by the model matches in most cases the observed two slip system scenario. A mathematical framework to predict the local Bragg angle based on the stretch and rotation of the crystal lattice by the elastic strain tensor was derived. Post-processing the intragranular strains fields from the CPFE results allowed to simulate 3D rocking curves, showing excellent agreement with the experimental measurements. This result confirms that in situ X-ray topotomography is a promising tool to study the early stage of polycrystal plasticity within the bulk of millimetric material specimens.

Made available in DSpace on 2015-05-14T12:36:38Z (GMT). No. of bitstreams: 1 arquivototal.pdf: 2501696 bytes, checksum: a1f65e686060aa68aa5e43dee1e6c0b5 (MD5) Previous issue date: 2012-07-30 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / Reliable identification systems have become key components in many applications that provide services to authenticated users. Since traditional authentication methods (such as using passwords or smartcards) can be manipulated in order to bypass the systems, biometric authentication methods have been receiving more attention in recent years. One of the biometric traits is the face. The problem of recognizing faces in video and photo still is an object of research, since there are many factors that influence the detection and recognition, such as lighting, position of the face, the background image, different facial expressions, etc. One can perform face recognition using Discrete Cosine Transform (DCT). In order to adjust a face recognition system to uncontrolled environments, two improvements for it were developed in this work: a image normalization module with respect to rotation and scale, and a change in the feature extraction module through the insertion of a non-ideal low-pass filter. The system and its modifications were tested on the following face databases: UFPB, ORL, Yale, and VSoft GTAV, developed specially for the job. Tests showed the efficiency of the image normalization module, but the system still is not adequate for every environment. / Sistemas de identificação confiáveis tornaram-se componentes chaves de várias aplicações que disponibilizam serviços para usuários autenticados. Uma vez que métodos de autenticação tradicionais (como os que utilizam senhas ou smartcards) podem ser manipulados com o objetivo de burlar os sistemas, métodos de autenticação biométrica vêm recebendo mais atenção nos últimos anos. Um dos traços biométricos é a face. O problema do reconhecimento de faces em vídeo e foto é objeto de pesquisa, uma vez que existem muitos fatores que influenciam na detecção e no reconhecimento, tais como: iluminação, posição da face, imagem ao fundo, diferentes expressões faciais, etc. É possível realizar reconhecimento facial utilizando a Transformada Cosseno Discreta (DCT). Com o intuito de adequar um Sistema de Detecção e Reconhecimento de Faces a ambientes não controlados, neste trabalho foram desenvolvidas duas melhorias para ele: um módulo normalizador de imagens em relação à rotação e à escala e uma modificação na etapa de seleção de atributos, por meio da inserção de um filtro passa-baixas não ideal. O sistema e suas modificações foram testados nos bancos de faces UFPB, ORL, Yale, GTAV e Vsoft, desenvolvido especialmente para o trabalho. Os testes mostraram a eficácia do módulo de normalização da imagem, mas ainda assim o sistema não é adequado para qualquer ambiente.

Millions of pictures are captured each year for different purposes, making digital images an ubiquitous part of modern day life. This proliferation was made possible by image compression standards since these images need to be stored somewhere and somehow. In this thesis I explore the use of machine learning together with the discrete cosine transform to compress images. An autoencoder was developed which was able to compress images with results comparable to the JPEG standard. The results lend credence to the hypothesis that the combination of a simple autoencoder and the discrete cosine transform offers a simple and effective method for image compression.