Mozzarella cheese is currently the highest consumed cheese in the United States. The popularity of mozzarella cheese is typically attributed to the high consumption rates of pizza cheese and string cheese; both of which are low moisture part skim (LMPS) mozzarella cheese. A single serving of LMPS mozzarella cheese contains approximately 8% of the daily value (DV) for sodium, a mineral which is currently consumed in excess among most Americans. On average, one in three Americans has hypertension. This condition is strongly associated with excessive sodium intake, and it is a leading risk factor for cardiovascular disease and stroke in the United States. Considering the popularity of LMPS mozzarella cheese, its sodium content, and the alarmingly high rates of hypertension among the American population, mozzarella cheese appears to be a product worth pursuing for sodium reduction. Salt (NaCl) provides several key benefits to cheese including: flavor enhancement, preservation, moisture control, and syneresis; thus reducing its concentration in cheese can result in multiple quality concerns. Previous research has investigated the effects of lower sodium in a variety of cheeses including: cheddar, feta, and imitation cheese. Typical methods for reducing sodium content in cheese include reduction of NaCl alone to a level which is still acceptable or partial substitution of NaCl with salt replacers including KCl. For assessing the consumer acceptability of low sodium cheeses, researchers have typically employed the use of traditional hedonic, preference, and ranking questions; however, consumer scientists have recently suggested the benefits of asking consumers questions which go beyond typical acceptability questions. Purchase intent and decision making have been associated with consumer emotions, and perhaps by better understanding consumer emotions toward nutritional alternative foods, including low sodium mozzarella cheese, a more successful low sodium cheese can be developed. The current study implemented a series of tests to progressively understand the role NaCl plays in mozzarella cheese and consumer opinion of low sodium cheese. A series of traditional sensory tests, including triangle, duo-trio, and hedonic tests, were performed to determine a sensory transparent antimicrobial and a consumer acceptable salt replacer. Based upon this sequence of tests, a single antimicrobial (SEA-i F75) and concentration (0.275%) were selected in addition to the most consistently preferred salt replacer in a low sodium mozzarella cheese system; however, the most preferred salt replacer was dependent upon the type of mozzarella cheese (direct acid or bacteria cultured). In order to address the potentially limited information from traditional sensory testing, a novel method for evaluating emotions was developed. The IMET (Image Measurement of Emotion and Texture) method utilized consumer provided images of emotions, researcher generated emotion images, and emotion words (the current industry standard) to aid in emotion testing, and the use of texture images and texture words (industry standard) for texture assessment. The IMET method was tested and validated across three commercial food product categories: orange soda, dairy beverages, and convenience cheeses. The IMET study indicated consumer selected emotion images were less variable than emotion words in a positive emotion, but the words only method was less variable in a negative emotion. Additionally, subjects were more likely to use images of themselves for positive emotions, and images of others for the negative emotions. After validation of the IMET method, the consumer emotion images method was used in conjunction with consumer acceptability testing and instrumental texture analysis in non-commercial low sodium mozzarella cheese. This study indicated the full sodium cheese was consistently liked most, followed by the 100% KCl cheese sample. Additionally, cheese with higher hedonic scores had increased changes in the positive emotions, while the disliked products had increased changes in the negative emotions across the eating experience. The instrumental texture analysis resulted in significant textural differences between the eight samples tested, and samples with higher mean scores for all texture attributes were associated with having higher mean hedonic scores. The studies performed in this thesis are important contributions for better understanding the role of NaCl in LMPS mozzarella cheese, and the consumer’s perception and potential acceptance of this nutritional alternative product. Additionally, the development of a novel emotion testing method may impact how researchers ask consumer questions, conduct consumer research, and investigate the effects of images on emotion testing with consumers.

Effective management of forest resources requires reliable and timely information on their status. In this regard, remote sensing techniques have played an important role, as they allow collection of data on extensive, remote and inaccessible areas. Historically, aerial photographs were the primary remote sensing data source in forest inventory and mapping, and they are still extensively for visual photo-interpretation. In this thesis, we show that their use can be improved thanks to automatic processing and an application using digitised orthophotos is provided. Satellite-based remote sensing has been regarded as an alternative, low-cost and rapid, data source to aerial photography and ground survey. Indeed, it has proved to be effective at the continental and global scales, but applications for local forest management purposes are still rare. The main reason for this is that the spatial resolution of satellite remote sensing data that was available until recently (mainly from Landsat TM/ETM and SPOT HRV) was too coarse for stand level information. Satellite images with enhanced spatial resolution (such as IKONOS) should overcome this limitation. This thesis investigates their actual capabilities for forest stand mapping and characterisation. We show that they are well suited for forest stand type classification and for retrieval of several dendrometric variables in coniferous stands with an accuracy similar to that of field sampling. For the sake of solutions to provide more precise and detailed information on forest stand, we assessed also the contribution of hyperspectral and multiple-view angle data acquired by CHRIS/PROBA. Although the winter season scene did not fully permit utilisation of the hyperspectral dimension of this dataset, the study provides insights into directional effects. This work makes, hopefully, a step towards automated processing and effective integration of satellite-based remote sensing data into the forest management information system and, by upscaling, into the national forest inventories.

In west-central Alberta increased landscape fragmentation has lead to increased human use, having negative effects on wildlife such as the grizzly bear (<i>Ursus arctos</i> L.). Recently, grizzly bears in the Foothills Model Forest were found to select clear cuts of different age ranges as habitat and selected or avoided certain clear cuts depending on the site preparation process employed. Satellite remote sensing offers a practical and cost-effective method by which cut areas, their age, and site preparation activities can be quantified. This thesis examines the utility of spectral reflectance of SPOT-5 pansharpened imagery (2.5m spatial resolution) to identify and map 44 regenerating stands sampled in August 2005. Using object based classification with the Normalized Difference Moisture Index (NDMI), green, and short wave infrared (SWIR) bands, 90% accuracy can be achieved in the detection of forest disturbance. Forest structural parameters were used to calculate the structural complexity index (SCI), the first loading of a principal components analysis. The NDMI, first-order standard deviation and second-order correlation texture measures were better able to explain differences in SCI among the 44 forest stands (R2=0.74). The best window size for the texture measures was 5x5, indicating that this is a measure only detectable at a very high spatial resolution. Age classes of these cut blocks were analysed using linear discriminant analysis and best separated (82.5%) with the SWIR and green spectral bands, second order correlation under a 25x25 window, and the predicted SCI. Site preparation was best classified (90.9%) using the NDMI and homogeneity texture under a 5x5 window. Future applications from this research include the selection of high probability grizzly habitat for high spatial resolution imagery acquisition for detailed mapping initiatives.

In west-central Alberta increased landscape fragmentation has lead to increased human use, having negative effects on wildlife such as the grizzly bear (<i>Ursus arctos</i> L.). Recently, grizzly bears in the Foothills Model Forest were found to select clear cuts of different age ranges as habitat and selected or avoided certain clear cuts depending on the site preparation process employed. Satellite remote sensing offers a practical and cost-effective method by which cut areas, their age, and site preparation activities can be quantified. This thesis examines the utility of spectral reflectance of SPOT-5 pansharpened imagery (2.5m spatial resolution) to identify and map 44 regenerating stands sampled in August 2005. Using object based classification with the Normalized Difference Moisture Index (NDMI), green, and short wave infrared (SWIR) bands, 90% accuracy can be achieved in the detection of forest disturbance. Forest structural parameters were used to calculate the structural complexity index (SCI), the first loading of a principal components analysis. The NDMI, first-order standard deviation and second-order correlation texture measures were better able to explain differences in SCI among the 44 forest stands (R2=0.74). The best window size for the texture measures was 5x5, indicating that this is a measure only detectable at a very high spatial resolution. Age classes of these cut blocks were analysed using linear discriminant analysis and best separated (82.5%) with the SWIR and green spectral bands, second order correlation under a 25x25 window, and the predicted SCI. Site preparation was best classified (90.9%) using the NDMI and homogeneity texture under a 5x5 window. Future applications from this research include the selection of high probability grizzly habitat for high spatial resolution imagery acquisition for detailed mapping initiatives.

In this thesis we develop algorithms for the processing of textures and images using a ladder-based biorthogonal directional filter bank (DFB). This work is based on the DFB originally proposed by Bamberger and Smith. First we present a novel implementation of this filter bank using ladder structures. This new DFB provides non-trivial FIR perfect reconstruction systems which are computationally very efficient. Furthermore we address the lack of shift-invariance in the DFB by presenting a novel undecimated DFB that preserves the computational simplicity of its maximally decimated counterpart. Finally, we study the use of the DFB in combination with pyramidal structures to form polar-separable image decompositions. Using the proposed filter banks we develop and evaluate algorithms for texture classification, segmentation and synthesis. We perform a comparative study with other image representations and find that the DFB provides some of the best results reported on the data sets used. Using the proposed directional pyramids we adapt wavelet thresholding algorithms. We find that our decompositions provide better edge and contour preservation than the best results reported using the undecimated discrete wavelet transform. Finally, we apply the developed algorithms to the analysis and processing of synthetic aperture radar (SAR) imagery. SAR image analysis is impaired by the presence of speckle noise. Our first objective will be to study the removal of speckle to enhance the visual quality of the image. Additionally, we implement land cover segmentation and classification algorithms taking advantage of the textural characteristics of SAR images. Finally, we propose a model-based SAR image compression algorithm in which the speckle component is separated from the structural features of a scene. The speckle component is captured with a texture model and the scene component is coded with a wavelet coder at very low bit rates. The resulting decompressed images have a better perceptual quality than SAR images compressed without removing speckle.

In this thesis, two different approaches are used to extract the relevant features for classifying the aflatoxin contaminated and uncontaminated scaled chili pepper samples: Statistical approach and Local Discriminant Bases (LDB) approach. In the statistical approach, First Order Statistical (FOS) features and Gray Level Cooccurrence Matrix (GLCM) features are extracted. In the LDB approach, the original LDB algorithm is modified to perform 2D searches to extract the most discriminative features from the hyperspectral images by removing irrelevant features and/or combining the features that do not provide sufficient discriminative information on their own. The classification is performed by using Linear Discriminant Analysis (LDA) classifier. Hyperspectral images of scaled chili peppers purchased from various locations in Turkey are used in this study. Correct classification accuracy about 80% is obtained by using the extracted features.

The objective of this study was to examine the link between radiographic measures of bone quality and total knee implant migration as measured by radiostereometric analysis (RSA). Two uncemented total knee arthroplasty studies (n=65) with RSA and bone mineral density (BMD) exams up to two years post surgery, and one study with cemented total knees with one year RSA data (n=18) were examined. Radiograph image texture analysis was used to characterize the bone microarchitecture, and a feasibility study was conducted to determine if a given x-ray machine could be used to obtain bone mineral density at the same time as the RSA exams. Random ForestTM ensemble classification tree statistical models classified patients into groups based on implant migration with a range of cut-points. Models based on bone texture parameters measured from the two year radiographs had a sensitivity of 87.5% and specificity of 80% when classifying patients who had more than 0.3mm maximum total point motion (MTPM) at two years using the one year exam as reference. Other cut-points were examined, with models generally having a lower specificity if the acceptable migration was smaller, and lower sensitivity if higher migrations were tolerable. In a predictive model, post-operative bone texture could be used to create a model with a sensitivity of 75% and a specificity of 80% when predicting those subjects with cemented implants who went on to more than 0.4mm total migration by one year. Bone mineral density of the proximal tibia, as determined by clinical scanners, was not found to increase the accuracy of implant migration group classification. An empirical fit to central regions of a purposed-built cross-wedge calibration phantom returned residuals of less than ±1.5% for the bone-equivalent thicknesses. The coefficient of variation of the region greyscale values in three images spread over three days is under 4%, showing the stability of the system to hold a calibration between phantom exams and patient scans. Scatter and dynamic range issues will need to be considered for an accurate calibration across the full range of areal bone mineral densities in the distal femur and proximal tibia.

nÃo hà / A anÃlise de imagens à um importante campo da visÃo computacional cujo propÃsito à extrair informaÃÃes significativas de imagens. Entre os vÃrios atributos relevantes que podem ser analisados, a textura à um dos mais importantes por ser uma fonte rica de informaÃÃes. O objetivo desta tese à desenvolver novos mÃtodos de anÃlise de textura (nÃveis de cinza e coloridas) baseados em modelos gravitacionais simplicados e caminhos mais curtos em grafos, que propiciem vetores de caracterÃsticas mais discriminativos do que os mÃtodos jà estabelecidos pela literatura. A primeira abordagem converte uma imagem em um sistema gravitacional simplificado cujo processo de colapso à explorado por meio de descritores de dimensÃo fractal e lacunaridade. A segunda abordagem converte os pixels de uma imagem em vÃrtices de um grafo ponderado nÃo-orientado e explora os caminhos mais curtos entre pares de vÃrtices em diferentes escalas e orientaÃÃes. Adicionalmente, nesta tese à proposto o estudo dessas abordagens na anÃlise de imagens de folhas de plantas para facilitar o moroso processo de taxonomia vegetal (problema este especialmente relevante para os botÃnicos) e de imagens mÃdicas para identificaÃÃo/classificaÃÃo de patologias, auxiliando o diagnÃstico mÃdico. Os experimentos sÃo realizados nas bases de imagens: Brodatz, UIUC, VisTex, USPTex, Outex, texturas foliares, parÃnquima paliÃÃdico, pap-smear e de tecido mamÃrio. Os resultados mais significativos de classificaÃÃo sÃo obtidos das bases UIUC, USPTex e parÃnquima paliÃÃdico, com taxas de acertos de 55,00%, 96,57% e 91,56% (menores taxas) obtidas pelos mÃtodos propostos, respectivamente. Essas taxas de acertos sÃo quase sempre superiores aos resultados obtidos pelos mÃtodos usados para comparaÃÃo, demonstrando que os mÃtodos propostos abrem promissoras fontes de pesquisa para os estudos de anÃlise de texturas em nÃveis de cinza e coloridas. / Image analysis is an important field of computer vision whose role is to extract significant information from images. Among several relevant attributes, texture is one of the most important because it is a rich source of information. This thesis aims to develop novel texture analysis methods (for grayscale and color images) based on simplified gravitational systems and shortest paths in graphs which provide feature vectors more discriminative than the methods already established in literature. The first approach converts an image into a simplified gravitational system whose collapse process is explored by using fractal dimension and lacunarity descriptors. The second approach converts the pixels of an image into vertices of a non-oriented weighted graph and explores the shortest paths between pairs of vertices in different scales and orientations. Additionally, this thesis proposes to apply these approaches to plant leaf identification (a relevant problem for botanists), and medical image identification/classication, increasing the confidence of medical diagnosis. The experiments are performed on the following image databases: Brodatz,UIUC, VisTex, USPTex, Outex, leaf textures, palisade parenchyma, pap-smear and breast tissues. The most significant comparison results are obtained from UIUC, USPTex and palisade parenchyma, with success rates of 55,00%, 96,57% and 91,56% (lower success rates) obtained by the proposed methods, respectively. These success rates are almost always superior to the results obtained by the methods used for comparison. This demonstrates that the proposed methods open promising sources of research in grayscale and color texture analysis.

Orientadores: Konradin Metze, Carlos Lenz Cesar / Tese (doutorado) - Universidade Estadual de Campinas, Faculdade de Ciências Médicas / Made available in DSpace on 2018-08-21T10:46:07Z (GMT). No. of bitstreams: 1 Vieira-Damiani_Gislaine_D.pdf: 4044155 bytes, checksum: 7a862a3866981d1827845318ce37b121 (MD5) Previous issue date: 2012 / Resumo: A hipertensão arterial sistêmica (HAS) bem como o envelhecimento provoca mudanças na estrutura dos grandes vasos sanguíneos - aorta e seus ramos - propiciando o desenvolvimento de processos degenerativos que são a causa de diversas doenças. O uso de ferramentas fotônicas na aquisição de imagens, associado a recursos matemáticos para a interpretação delas representa um avanço para as análises histopatológicas, pois permitem a visualização e compreensão de pequenas estruturas que antes eram impossíveis de serem observadas. O objetivo desse trabalho foi associar estas duas tecnologias (ferramentas fotônicas e recursos matemáticos) e com isso criar uma metodologia para a análise simultânea de fibras elásticas e colágenas na aorta. Para tanto utilizamos aorta ascendente de 72 pacientes, sendo 22 normotensos, 38 portadores de HAS e 12 aortas de dissecção. As lâminas coradas com hematoxilina eosina foram examinadas no microscópio multifoton, com dois fótons: laser de argônio para fluorescência da eosina, corante de fibras elásticas e Ti:safira para SHG, sinal gerado por moléculas de colágeno. A distribuição e organização das fibras elásticas e colágenas foram analisadas pelas seguintes variáveis: morfometria geométrica, derivadas da matriz de co-ocorrência de Haralick, Transformada de Fourier e fluorescência ótica integrada. Usando estes descritores da textura associados a fractais, observamos que a geração do SHG é dependente não só da presença do colágeno como também do arranjo destas fibras. Observamos ainda que em indivíduos normotensos, quando comparados aos portadores de HAS, ocorre uma diminuição na distribuição do sinal SHG ao longo da espessura da camada média partindo da íntima em sentido à adventícia. Dessa maneira concluímos que os maiores distúrbios das fibras elásticas, nos indivíduos normais ocorrem na transição do terço interno para o médio, enquanto que nos portadores de HAS eles estão distribuídos em toda a espessura da aorta. Além disso, estes estudos nos permitiram verificar que a dissecção da aorta ocorre entre dois reforços de colágeno, uma vez que este fenômeno foi constatado entre dois picos de SHG / Abstract: The arterial hypertension as well as aging induces changes in the structure of large blood vessels - aorta and its branches - leading to development of degenerative processes which are the cause of many diseases. The use of photonics tools for image acquisition, associated to mathematical resources for interpretation of them represents an advance in histopathological analysis, because it allows the visualization and understanding of small structures that were impossible to be observed before. The main objective of this study was to associate both technologies (photonics tool and mathematical resources) to create a new methodology to evaluate, simultaneously, elastic and collagen fibers in aorta. For this we've used autopsies of ascending aortas from 72 patients, being 22 samples from normotensives individuals, 38 from HAS patients and 12 aortas from dissection. HE-stained paraffin sections from ascending aortas were analyzed by multifoton microscopy, with 2 types of photons: Two-photon excited fluorescence (TPEF) for elastin and Ti:safira for SHG to analyze collagen fibers. The distribution and organization of elastic and collagen fibers were analyzed by the following variables: geometric morphometric, derived from the co-occurrence matrix of Haralick, Fourier Transform and Fluorescence optics integrated. Using these texture descriptors associated to analysis of fractals, we've observed that SHG generation is not only dependent on the presence of collagen but on the arrangement of these fibers as well. We also observed that in normotensives individuals, if compared to HAS patients, occurs a decrease in the SHG intensity along the medial thickness from intimate in direction to adventitia. Thus we conclude that the major disorders of elastic fibers in normal subjects occur in the transition from the third layer to the middle, while in HAS individuals these disorders are distributed throughout the thickness of the aorta. Furthermore, this study has allowed us to verify that the aortic dissection has occurred between two peaks of SHG, since this phenomenon was observed between two ribs collagen / Doutorado / Biologia Estrutural, Celular, Molecular e do Desenvolvimento / Doutora em Fisiopatologia Médica

Dans le contexte actuel de diminution des pollutions d’origine agricole, laréduction des apports d’intrants devient un enjeu primordial. En France, laviticulture est l’activité qui possède le taux le plus important de traitementsphytosanitaires par unité de surface. Elle représente, à elle seule, 20% de laconsommation annuelle de pesticides. Par conséquent, il est nécessaire d’étudierle devenir des pesticides appliqués afin de réduire les quantités perduesdans l’environnement. Dans le cadre de la réduction d’apport de produitsphytosanitaires dans les vignes, de nombreux travaux ont été effectués sur lamodélisation du comportement d’un spray de gouttelettes et sa répartitionau niveau de la parcelle et de l’air environnant. Cependant, il est égalementimportant de s’intéresser au comportement de la gouttelette directement auniveau de la feuille. Les progrès dans le domaine de l’imagerie et la diminutiondu coût des systèmes ont rendus ces systèmes beaucoup plus attractifs.Le travail de cette thèse consiste en la mise en place d’un système d’imagerierapide qui permet l’observation du comportement à l’impact de gouttelettesrépondant aux conditions de pulvérisation. Les caractéristiques ainsi que lecomportement associé de chaque gouttelette sont extraits grâce à une méthodede suivi d’objets. Une analyse statistique basée sur un nombre représentatifde résultats permet ensuite d’évaluer de manière robuste le devenir d’unegoutte en fonction de ses caractéristiques. Parallèlement, un paramètre décrivantl’état de surface de la feuille est également étudié grâce à l’imagerie : larugosité qui joue un rôle prédominant dans la compréhension des mécanismesd’adhésion / In the domain of vineyard precision spraying research, one of the most importantobjectives is to minimize the volume of phytosanitary products ejected bya sprayer in order to be more environmentally respectful with more effectivevine leaf treatments. Unfortunaltely, even if lot of works have been carriedout at a parcel scale, mainly on losses caused by drift, less works have beencarried out at the leaf scale in order to understand which parameters influencethe spray quality. Since few years, recent improvements in image processing,sensitivity of imaging systems and cost reduction have increased the interestof high-speed imaging techniques. Analyzing the behavior of droplets afterimpact with the leaf thanks to high speed imaging technology is a relevantsolution. By this way, we propose a droplets behavior analyzing process invineyard spraying context based on high-speed acquision system combinedwith image processing techniques. This process allows us to extract dropletsparameters. Therefore, a statistical study is processed in order to determinethe effects of droplets parameters on leaf impact or to predict behavior of asingle droplet. Since this behavior is strongly related to leaf surface, we alsopropose to validate a natural leaf roughness characterization method basedon texture analysis