No description available.

Ce mémoire aborde le problème de la détection de visages à partir d'une image compressée. Il touche également à un problème connexe qui est la qualité des standards de compression et l'estimation de celle-ci. Ce mémoire est organisé sous la forme d'une introduction générale sur la détection de visages et de deux articles soumis à des conférences internationales. Le premier article propose une amélioration de la méthode classique pour comparer la qualité de deux standards. Le deuxième propose une méthode de décompression spécialisée pour faire fonctionner le détecteur de visages de Viola-Jones dans le domaine compressé.

The validity of the speech act data obtained from different types of elicitation instruments has been widely debated in the Interlanguage Pragmatics literature. This study compared the use of apology strategies and modality markers from apology speech act data collected from two most popular speech act elicitation tasks, namely, Discourse Completion Task (DCT) and role plays. Sixty native speakers of Chinese (NS-C) and sixty EFL learners responded to DCT. Twenty four NS-C and forty EFL learners participated in role plays. Results show that subjects tended to use Direct Expression of Apology, Acknowledge Responsibility and Offer of Repairs as their main strategies in apology situations both on DCT and in role plays. In addition, participants tended to exploit MAXIMIZERs more than MINIMIZERs in both methods. However, differences between two approaches were revealed when frequencies and distributions of apology strategies and modality markers were examined. Role plays elicited overall more apology strategies and modality markers than DCT did. Participants were also found to employ a narrower range of strategies and modality markers on DCT. Because of the feeling of insecurity in face-to-face encounters (Rintell and Mitchell, 1989), EFL-L exhibited more ¡¥play-it-safe¡¦ strategies (Faerch and Kasper, 1989; Trosborg, 1987) by giving more direct apologies, and exploiting more MAXIMIZERs than MINIMIZERs across four situations. When dividing NS-C into two separate groups: English majors (NS-C-EM) and non-English (NS-C-NEM), some in-group differences were revealed. For instance, in role-play data, the preference order for apology strategy choice was different between NS-C-EM and NS-C-NEM. Also, NS-C-EM tended to exploit overall more Chinese particles, a, ba, and ne, than their NS-C-NEM counterparts. The finding suggests that it is necessary to divide NS-C into different groups based on their educational background such as English majors and non-English majors. The present study also pointed out that traditional categorization of apology internal modification which was based on the language system of English may fail to fully capture Chinese apology behaviors. Chinese modal particles which have no English equivalent (Tang and Tang, 1997) have been ignored in the categorization of apology internal modifications in the literature. Thus, a modified coding system which included Chinese particles, namely, ¡§A¡¨ (°Ú), ¡§BA¡¨ (§a), and ¡§NE¡¨ (©O), was proposed based on traditional categorization of apology internal modification developed in the previous studies (Blum-Kulka & Kasper, 1989; Lin and Ho, 2006; Trosborg, 1995). The result indicates that while the previous study (Lin and Ho, 2006) on apology internal modification which excluded Chinese particles has revealed that NS-C exploited less modality markers than their native speakers of English (NS-E) counterparts, the result in the present showed an opposite pattern in that by tagging Chinese particles at the end of the utterances, NS-C were found to modulate their tone more often than NS-E. This might imply the importance of the Chinese particles for NS-C in conveying attitude. Finally, in order to increase the validity of data elicitation methods, further studies addressed to the methodological issue should include the analysis of the responses obtained from naturally occurring data and examine whether both data obtained from DCT and role plays are representative of ¡¥natural speech¡¦ . Also, in order to have cross-cultural and cross-linguistic comparison, speech acts data produced by NS-E, NS-C, and EFL learners collected through DCT, role plays, and naturally occurring data are needed for future researches.

Digital image and video in their raw form require an enormous amount of storage capacity. Considering the important role played by digital imaging and video, it is necessary to develop a system that produces high degree of compression while preserving critical image/video information. There are various transformation techniques used for data compression. Discrete Cosine Transform (DCT) and Discrete Wavelet Transform (DWT) are the most commonly used transformation. DCT has high energy compaction property and requires less computational resources. On the other hand, DWT is multiresolution transformation.<p> In this work, we propose a hybrid DWT-DCT algorithm for image compression and reconstruction taking benefit from the advantages of both algorithms. The algorithm performs the Discrete Cosine Transform (DCT) on the Discrete Wavelet Transform (DWT) coefficients. Simulations have been conducted on several natural, benchmark, medical and endoscopic images. Several QCIF, high definition, and endoscopic videos have also been used to demonstrate the advantage of the proposed scheme.<p> The simulation results show that the proposed hybrid DWT-DCT algorithm performs much better than the standalone JPEG-based DCT, DWT, and WHT algorithms in terms of peak signal to noise ratio (PSNR), as well as visual perception at higher compression ratio. The new scheme reduces false contouring and blocking artifacts significantly. The rate distortion analysis shows that for a fixed level of distortion, the number of bits required to transmit the hybrid coefficients would be less than those required for other schemes Furthermore, the proposed algorithm is also compared with the some existing hybrid algorithms. The comparison results show that, the proposed hybrid algorithm has better performance and reconstruction quality. The proposed scheme is intended to be used as the image/video compressor engine in imaging and video applications.

Digital image and video in their raw form require an enormous amount of storage capacity. Considering the important role played by digital imaging and video, it is necessary to develop a system that produces high degree of compression while preserving critical image/video information. There are various transformation techniques used for data compression. Discrete Cosine Transform (DCT) and Discrete Wavelet Transform (DWT) are the most commonly used transformation. DCT has high energy compaction property and requires less computational resources. On the other hand, DWT is multiresolution transformation.<p> In this work, we propose a hybrid DWT-DCT algorithm for image compression and reconstruction taking benefit from the advantages of both algorithms. The algorithm performs the Discrete Cosine Transform (DCT) on the Discrete Wavelet Transform (DWT) coefficients. Simulations have been conducted on several natural, benchmark, medical and endoscopic images. Several QCIF, high definition, and endoscopic videos have also been used to demonstrate the advantage of the proposed scheme.<p> The simulation results show that the proposed hybrid DWT-DCT algorithm performs much better than the standalone JPEG-based DCT, DWT, and WHT algorithms in terms of peak signal to noise ratio (PSNR), as well as visual perception at higher compression ratio. The new scheme reduces false contouring and blocking artifacts significantly. The rate distortion analysis shows that for a fixed level of distortion, the number of bits required to transmit the hybrid coefficients would be less than those required for other schemes Furthermore, the proposed algorithm is also compared with the some existing hybrid algorithms. The comparison results show that, the proposed hybrid algorithm has better performance and reconstruction quality. The proposed scheme is intended to be used as the image/video compressor engine in imaging and video applications.

The first topic of this thesis is a novel voltage tripler using 4 clocks with different phases. Both the positive and negative polarities of the voltage are generated to serve as the boosted voltage and the back bias voltage. The proposed design is carried out by pass transistors and switched capacitors. The second topic is a low-power discrete cosine transform (DCT) processor. It is suitable for portable applications. The number of clock cycles needed for processing an 8¡Ñ8 block of pixels is increased, but the chip area is reduced. The reduction of the chip area leads to the reduction of the power dissipation.

This study aimed to investigate Chinese EFL interlanguage complaint behaviors in terms of production collected by role play and DCT. Four contextual factors were involved: social distance, social status, gender and obligation. A total of 320 role plays were elicited from 80 college students, including 20 native speakers of English, 20 native speakers of Chinese, 20 EFL-low proficiency learners and 20 EFL-high proficiency learners. Additionally, the DCT data were elicited by 180 participants, including 60 native speakers of Chinese, 60 native speakers of English, 30 EFL-high proficiency learners and 30 EFL-low proficiency learners. Three aspects of complaints were examined: complaint strategies, complaint modifications and the effect of language proficiency. Results showed that no matter in role play or on DCT, participants tended to use more indirect strategies to avoid conflicts with others. Below the level of Reproach was the main complaint strategy they used. Besides, the complaint strategies used by the four groups were quite similar in role play. However, they differed in the frequency of strategy use, the preference orders, and modifications used in making their complaints. On the other hand, participants performed quite differently on DCT. They differed in strategy use, but there were no significant difference in frequency use, preference orders, and the use of modifications. Furthermore, NS-E tended to use more direct strategies than EFL learners. This result echoed Olshtain and Weinbach¡¦s (1993) findings that learners would use less severe strategies than native speakers because learners would try to ¡§¡Kavoid straightforward face-threatening interactions at all costs¡¨ (p. 115). Besides, proficiency effect is found in EFL learners¡¦ interlanguage complaints production. It is more difficult for EFL-L learners to make complaints because of their low L2 proficiency. For future studies, gathering natural occurring data to compare the differences between role play is suggested. Besides, due to time limitation, the issue of interlocutor gender differences is not discussed and future work could include this matter to provide a more comprehensive understanding of the speech at of complaints.

Submitted by João Arthur Martins (joao.arthur@ufpe.br) on 2015-03-10T19:14:48Z No. of bitstreams: 2 DISSERTAÇÃO Hugo Albuquerque.pdf: 7422850 bytes, checksum: 3133c927702ab38bd7ee89b9c3f5cc73 (MD5) license_rdf: 1232 bytes, checksum: 66e71c371cc565284e70f40736c94386 (MD5) / Approved for entry into archive by Daniella Sodre (daniella.sodre@ufpe.br) on 2015-03-10T19:43:36Z (GMT) No. of bitstreams: 2 DISSERTAÇÃO Hugo Albuquerque.pdf: 7422850 bytes, checksum: 3133c927702ab38bd7ee89b9c3f5cc73 (MD5) license_rdf: 1232 bytes, checksum: 66e71c371cc565284e70f40736c94386 (MD5) / Made available in DSpace on 2015-03-10T19:43:36Z (GMT). No. of bitstreams: 2 DISSERTAÇÃO Hugo Albuquerque.pdf: 7422850 bytes, checksum: 3133c927702ab38bd7ee89b9c3f5cc73 (MD5) license_rdf: 1232 bytes, checksum: 66e71c371cc565284e70f40736c94386 (MD5) Previous issue date: 2013-08-20 / A fus~ao de imagens tem-se destacado na area de processamento de imagens, atrav es da ideia de combinar duas ou mais imagens que apresentem caracter sticas distintas (provenientes de diferentes sensores) ou alguma degrada c~ao (regi~oes desfocadas, mal iluminadas, com baixo contraste). O foco e uma propriedade intimamente relacionada com a qualidade da imagem. Em algumas situa c~oes, n~ao e poss vel manter simultaneamente todas as regi~oes da paisagem bem focadas. Uma alternativa e utilizar a fus~ao de imagens para combinar os focos diferentes de uma foto em apenas uma imagem com todas as regi~oes melhor focadas. Este trabalho apresenta uma ideia de utilizar diferentes t ecnicas de fus~ao de imagens em conjunto para compor o resultado nal. Para tanto, as imagens de entrada s~ao divididas em blocos, nos quais s~ao aplicados os algoritmos de fus~ao escolhidos. Em seguida, escolhe-se a t ecnica que atuou melhor em cada bloco, atrav es de uma m etrica de qualidade de imagem (como a PSNR (Peak Signal to Noise Ratio)), formando um mosaico com os melhores resultados. Dois algoritmos existentes na literatura envolvendo a fus~ao de imagens no dom nio da frequ^encia baseados em foco s~ao descritos: Contraste no dom nio da DCT (Transformada Discreta de Cosseno) e Frequ^encia Espacial. Modi ca c~oes s~ao feitas em ambas as t ecnicas a m de melhorar os crit erios de sele c~ao para cada bloco, por meio da de ni c~ao da escolha de um bloco inteiro ou dos pixels (ou coe cientes) individualmente. A qualidade da imagem nal e avaliada utilizando m etricas conhecidas, como a PSNR, RMSE (Root Mean Square Error), SSIM (Structural Similarity), entre outras. Tanto para as melhorias quanto para a combina c~ao de algoritmos, os resultados s~ao equivalentes ou superiores quando comparados aos das t ecnicas originais. Em caso de igualdade da qualidade, o algoritmo melhorado possui execu c~ao mais r apida que o original, justi cando assim o seu uso. O tempo de execu c~ao de cada algoritmo e avaliado, mostrando que as t ecnicas de combina c~ao aumentam o tempo de processamento, pois os algoritmos s~ao executados sequencialmente. Nesse caso, a utiliza c~ao de paralelismo pode acelerar a execu c~ao. Uma vez que a combina c~ao adiciona um processamento extra, estudos de viabilidade podem ser realizados para cada caso, a m de considerar a real necessidade do aumento da qualidade da imagem fundida, em detrimento dos recursos a serem utilizados.

Rapid evolution of digital technology has improved the ease of access to digital information enabling reliable, faster and efficient storage, transfer and processing of digital data. It also leads to the consequence of making the illegal production and redistribution of digital media easy and undetectable. Hence, the risk of copyright violation of multimedia data has increased due to the enormous growth of computer networks that provides fast and error free transmission of any unauthorized duplicate and possibly manipulated copy of multimedia information. One possible solution may be to embed a secondary signal or pattern into the image that is not perceivable and is mixed so well with the original digital data that it is inseparable and remains unaffected against any kind of multimedia signal processing. This embedded secondary information is digital watermark which is, in general, a visible or invisible identification code that may contain some information about the intended recipient, the lawful owner or author of the original data, its copyright etc. in the form of textual data or image. In order to be effective for copyright protection, digital watermark must be robust which are difficult to remove from the object in which they are embedded despite a variety of possible attacks. Several types of watermarking algorithms have been developed so far each of which has its own advantages and limitations. Among these, recently Singular Value Decomposition (SVD) based watermarking algorithms have attracted researchers due to its simplicity and some attractive mathematical properties of SVD. Here a number of pure and hybrid SVD based watermarking schemes have been investigated and finally a RST invariant modified SVD and Discrete Cosine Transform (DCT) based algorithm has been developed. A preprocessing step before the watermark extraction has been proposed which makes the algorithm resilient to geometric attack i.e. RST attack. Performance of this watermarking scheme has been analyzed by evaluating the robustness of the algorithm against geometric attack including rotation, scaling, translation (RST) and some other attacks. Experimental results have been compared with existing algorithm which seems to be promising. / Phone number: +88041730212

The main aim of this project is to implement two different types of the watermarking algorithm in frequency (transforamtion) domain. Because of the huge amount of the watermarking algorithms is this thesis dedicated to the watermarking methods in the frequency domain. These methods use kind of transformation implemented on the original data, the second step is modification of transformation coeficients. This project is dedicated to the methods which use discrete wavelet transformation. Nowdays the most popular transformation, which meets the best efforts. The aim of every method is to embeed the watermark into the original data so precisly, that none observer can recognize any change between the original and watermarked data. On the contrary, watermarked picture should be as resistant to the attacks as possible. By the term of attack, we understand all the ways, that disable watermark extraction. The watermark method robustness will be tested with the use of Checkmark program.