Thesis (M.A.)--Boston University / A method of reducing the graininess of a photographic print and increasing resolution in low contrast regions is described. The method involves the printing of more than one negative frame to produce one print. This requires a series of negatives with identical detail coverage in the area to be printed. The success of the method depends largely on the precision of the solution of the registration problem. Each negative is printed in turn, using the normal exposure, partitioned in as many parts as there are negatives to be printed. Each negative must be registered as exactly as possible in the image area. Four different aerial emulsions were used to obtain the 35-mm negatives for the superimposition printing technique. Kodak films used were: Tri X-RP Aercon, Super XX-RP Aerial Recon, Plus X Aerocon (SO 1166), and SO-1213. The exposure versus resolution characteristics and the basic sensitometric curves were developed for these films prior to exposure of the final series of negative frames. The negatives were exposed under identical conditions with the exception of lens openings and shutter speeds at an object to image ratio of 160 to 1. The camera was a Contax IIA with a 50-nm F/2 Sonnar lens. The camera exposure settings were: Tri-X, F/16, 1/250 second; Plus X, F/16, 1/100 second; SO-1213, F/11, 1/50 second. Due to the level of brightness of the target, the camera lens was not used at its best aperture. No filter was used. [TRUNCATED].

The ability to detect faces in images is of critical ecological significance. It is a pre-requisite for other important face perception tasks such as person identification, gender classification and affect analysis. Here we address the question of how the visual system classifies images into face and non-face patterns. We focus on face detection in impoverished images, which allow us to explore information thresholds required for different levels of performance. Our experimental results provide lower bounds on image resolution needed for reliable discrimination between face and non-face patterns and help characterize the nature of facial representations used by the visual system under degraded viewing conditions. Specifically, they enable an evaluation of the contribution of luminance contrast, image orientation and local context on face-detection performance.

No description available.

A segmentação de pele humana, em imagens coloridas, tem sido largamente estudada nos últimos anos servindo de fundamento para muitos outros estudos como, por exemplo, a detecção de faces. Dentre as inúmeras aplicações de trabalhos relativos à segmentação de pele humana está a de se localizar uma determinada pessoa em locais de grande concentração humana tais como: avenidas, terminais de ônibus, aeroportos, shopping centers e estádios. Desta forma, a necessidade de se obter um sistema que classifique de forma adequada a pele humana tornou-se a principal motivação para o desenvolvimento deste trabalho. Desta forma, propõe-se uma metodologia para melhorar a segmentação de pele humana em imagens coloridas através de um algoritmo mais eficiente. O algoritmo é baseado na média de vizinhanças cujos valores limites, para definição do intervalo de cor equivalente à pele humana, são obtidos através de uma imagem padrão, gerada a priori, com amostras de pele humana. Esta imagem é chamada de \"colcha de retalhos\". A metodologia tem como base de comparação trabalhos anteriores similares, principalmente o desenvolvido por Kovac et al. (2003). Os resultados mostram um desempenho superior da metodologia proposta. / The segmentation of human skin, in colored images, has been studied broadly for the last years serving as foundation for many other studies as, for instance, the detection of faces. Among the countless applications of works related to the segmentation of human skin it is the one of localizing a certain person in places of great human concentration such as: avenues, bus terminals, airports, shopping centers and stadiums. Therefore, the need to obtain a system that classifies in an appropriate way the human skin became the main motivation for the development of this work. This way, a methodology to improve the segmentation of human skin in colored images through a more efficient algorithm is proposed. The algorithm is based on the average of neighborhoods whose limit values, for definition of the interval of equivalent color to the human skins, are obtained through an image pattern, generated in priori, with samples of human skin. This image is called \"bedspread of remnants\". The methodology has as base of comparison similar previous works, mainly the one developed by Kovac et al. (2003). The results show a superior performance of the proposed methodology.

A segmentação de pele humana, em imagens coloridas, tem sido largamente estudada nos últimos anos servindo de fundamento para muitos outros estudos como, por exemplo, a detecção de faces. Dentre as inúmeras aplicações de trabalhos relativos à segmentação de pele humana está a de se localizar uma determinada pessoa em locais de grande concentração humana tais como: avenidas, terminais de ônibus, aeroportos, shopping centers e estádios. Desta forma, a necessidade de se obter um sistema que classifique de forma adequada a pele humana tornou-se a principal motivação para o desenvolvimento deste trabalho. Desta forma, propõe-se uma metodologia para melhorar a segmentação de pele humana em imagens coloridas através de um algoritmo mais eficiente. O algoritmo é baseado na média de vizinhanças cujos valores limites, para definição do intervalo de cor equivalente à pele humana, são obtidos através de uma imagem padrão, gerada a priori, com amostras de pele humana. Esta imagem é chamada de \"colcha de retalhos\". A metodologia tem como base de comparação trabalhos anteriores similares, principalmente o desenvolvido por Kovac et al. (2003). Os resultados mostram um desempenho superior da metodologia proposta. / The segmentation of human skin, in colored images, has been studied broadly for the last years serving as foundation for many other studies as, for instance, the detection of faces. Among the countless applications of works related to the segmentation of human skin it is the one of localizing a certain person in places of great human concentration such as: avenues, bus terminals, airports, shopping centers and stadiums. Therefore, the need to obtain a system that classifies in an appropriate way the human skin became the main motivation for the development of this work. This way, a methodology to improve the segmentation of human skin in colored images through a more efficient algorithm is proposed. The algorithm is based on the average of neighborhoods whose limit values, for definition of the interval of equivalent color to the human skins, are obtained through an image pattern, generated in priori, with samples of human skin. This image is called \"bedspread of remnants\". The methodology has as base of comparison similar previous works, mainly the one developed by Kovac et al. (2003). The results show a superior performance of the proposed methodology.

Image super-resolution aims to produce a higher resolution representation of a scene from an ensemble of low-resolution images that may be warped, aliased, blurred and degraded by noise. There are a variety of methods for performing super-resolution described in the literature, and in general they consist of three major steps: image registration, fusion and deblurring. This thesis proposes a novel method of performing the first two of these steps. The ultimate aim of image super-resolution is to produce a higher-quality image that is visually clearer, sharper and contains more detail than the individual input images. Machine algorithms can not assess images qualitatively and typically use a quantitative error criterion, often least-squares. This thesis aims to optimise leastsquares directly using a fast method, in particular one that can be implemented using linear filters; hence, a closed-form solution is required. The concepts of optimal interpolation and resampling are derived and demonstrated in practice. Optimal filters optimised on one image are shown to perform nearoptimally on other images, suggesting that common image features, such as stepedges, can be used to optimise a near-optimal filter without requiring the knowledge of the ground-truth output. This leads to the construction of a pulse model, which is used to derive filters for resampling non-uniformly sampled images that result from the fusion of registered input images. An experimental comparison shows that a 10th order pulse model-based filter outperforms a number of methods common in the literature. The use of optimal interpolation for image registration linearises an otherwise nonlinear problem, resulting in a direct solution. Experimental analysis is used to show that optimal interpolation-based registration outperforms a number of existing methods, both iterative and direct, at a range of noise levels and for both heavily aliased images and images with a limited degree of aliasing. The proposed method offers flexibility in terms of the size of the region of support, offering a good trade-off in terms of computational complexity and accuracy of registration. Together, optimal interpolation-based registration and fusion are shown to perform fast, direct and effective super-resolution.

Image super-resolution aims to produce a higher resolution representation of a scene from an ensemble of low-resolution images that may be warped, aliased, blurred and degraded by noise. There are a variety of methods for performing super-resolution described in the literature, and in general they consist of three major steps: image registration, fusion and deblurring. This thesis proposes a novel method of performing the first two of these steps. The ultimate aim of image super-resolution is to produce a higher-quality image that is visually clearer, sharper and contains more detail than the individual input images. Machine algorithms can not assess images qualitatively and typically use a quantitative error criterion, often least-squares. This thesis aims to optimise leastsquares directly using a fast method, in particular one that can be implemented using linear filters; hence, a closed-form solution is required. The concepts of optimal interpolation and resampling are derived and demonstrated in practice. Optimal filters optimised on one image are shown to perform nearoptimally on other images, suggesting that common image features, such as stepedges, can be used to optimise a near-optimal filter without requiring the knowledge of the ground-truth output. This leads to the construction of a pulse model, which is used to derive filters for resampling non-uniformly sampled images that result from the fusion of registered input images. An experimental comparison shows that a 10th order pulse model-based filter outperforms a number of methods common in the literature. The use of optimal interpolation for image registration linearises an otherwise nonlinear problem, resulting in a direct solution. Experimental analysis is used to show that optimal interpolation-based registration outperforms a number of existing methods, both iterative and direct, at a range of noise levels and for both heavily aliased images and images with a limited degree of aliasing. The proposed method offers flexibility in terms of the size of the region of support, offering a good trade-off in terms of computational complexity and accuracy of registration. Together, optimal interpolation-based registration and fusion are shown to perform fast, direct and effective super-resolution.

Image super-resolution aims to produce a higher resolution representation of a scene from an ensemble of low-resolution images that may be warped, aliased, blurred and degraded by noise. There are a variety of methods for performing super-resolution described in the literature, and in general they consist of three major steps: image registration, fusion and deblurring. This thesis proposes a novel method of performing the first two of these steps. The ultimate aim of image super-resolution is to produce a higher-quality image that is visually clearer, sharper and contains more detail than the individual input images. Machine algorithms can not assess images qualitatively and typically use a quantitative error criterion, often least-squares. This thesis aims to optimise leastsquares directly using a fast method, in particular one that can be implemented using linear filters; hence, a closed-form solution is required. The concepts of optimal interpolation and resampling are derived and demonstrated in practice. Optimal filters optimised on one image are shown to perform nearoptimally on other images, suggesting that common image features, such as stepedges, can be used to optimise a near-optimal filter without requiring the knowledge of the ground-truth output. This leads to the construction of a pulse model, which is used to derive filters for resampling non-uniformly sampled images that result from the fusion of registered input images. An experimental comparison shows that a 10th order pulse model-based filter outperforms a number of methods common in the literature. The use of optimal interpolation for image registration linearises an otherwise nonlinear problem, resulting in a direct solution. Experimental analysis is used to show that optimal interpolation-based registration outperforms a number of existing methods, both iterative and direct, at a range of noise levels and for both heavily aliased images and images with a limited degree of aliasing. The proposed method offers flexibility in terms of the size of the region of support, offering a good trade-off in terms of computational complexity and accuracy of registration. Together, optimal interpolation-based registration and fusion are shown to perform fast, direct and effective super-resolution.

Image super-resolution aims to produce a higher resolution representation of a scene from an ensemble of low-resolution images that may be warped, aliased, blurred and degraded by noise. There are a variety of methods for performing super-resolution described in the literature, and in general they consist of three major steps: image registration, fusion and deblurring. This thesis proposes a novel method of performing the first two of these steps. The ultimate aim of image super-resolution is to produce a higher-quality image that is visually clearer, sharper and contains more detail than the individual input images. Machine algorithms can not assess images qualitatively and typically use a quantitative error criterion, often least-squares. This thesis aims to optimise leastsquares directly using a fast method, in particular one that can be implemented using linear filters; hence, a closed-form solution is required. The concepts of optimal interpolation and resampling are derived and demonstrated in practice. Optimal filters optimised on one image are shown to perform nearoptimally on other images, suggesting that common image features, such as stepedges, can be used to optimise a near-optimal filter without requiring the knowledge of the ground-truth output. This leads to the construction of a pulse model, which is used to derive filters for resampling non-uniformly sampled images that result from the fusion of registered input images. An experimental comparison shows that a 10th order pulse model-based filter outperforms a number of methods common in the literature. The use of optimal interpolation for image registration linearises an otherwise nonlinear problem, resulting in a direct solution. Experimental analysis is used to show that optimal interpolation-based registration outperforms a number of existing methods, both iterative and direct, at a range of noise levels and for both heavily aliased images and images with a limited degree of aliasing. The proposed method offers flexibility in terms of the size of the region of support, offering a good trade-off in terms of computational complexity and accuracy of registration. Together, optimal interpolation-based registration and fusion are shown to perform fast, direct and effective super-resolution.

Image super-resolution aims to produce a higher resolution representation of a scene from an ensemble of low-resolution images that may be warped, aliased, blurred and degraded by noise. There are a variety of methods for performing super-resolution described in the literature, and in general they consist of three major steps: image registration, fusion and deblurring. This thesis proposes a novel method of performing the first two of these steps. The ultimate aim of image super-resolution is to produce a higher-quality image that is visually clearer, sharper and contains more detail than the individual input images. Machine algorithms can not assess images qualitatively and typically use a quantitative error criterion, often least-squares. This thesis aims to optimise leastsquares directly using a fast method, in particular one that can be implemented using linear filters; hence, a closed-form solution is required. The concepts of optimal interpolation and resampling are derived and demonstrated in practice. Optimal filters optimised on one image are shown to perform nearoptimally on other images, suggesting that common image features, such as stepedges, can be used to optimise a near-optimal filter without requiring the knowledge of the ground-truth output. This leads to the construction of a pulse model, which is used to derive filters for resampling non-uniformly sampled images that result from the fusion of registered input images. An experimental comparison shows that a 10th order pulse model-based filter outperforms a number of methods common in the literature. The use of optimal interpolation for image registration linearises an otherwise nonlinear problem, resulting in a direct solution. Experimental analysis is used to show that optimal interpolation-based registration outperforms a number of existing methods, both iterative and direct, at a range of noise levels and for both heavily aliased images and images with a limited degree of aliasing. The proposed method offers flexibility in terms of the size of the region of support, offering a good trade-off in terms of computational complexity and accuracy of registration. Together, optimal interpolation-based registration and fusion are shown to perform fast, direct and effective super-resolution.