�ݺ�ߣshows by User: GuillaumeGales

�ݺ�ߣshows by User: GuillaumeGales / http://www.slideshare.net/images/logo.gif �ݺ�ߣshows by User: GuillaumeGales / Thu, 28 Aug 2014 12:44:36 GMT �ݺ�ߣShare feed for �ݺ�ߣshows by User: GuillaumeGales Constant colour matting with foreground estimation /slideshow/imvip2014/38461321 imvip2014-140828124436-phpapp02
Constant colour matting consists of estimating for each pixel of an image the propor- tion α of an unknown foreground colour with a known constant background colour. The α-matte is then used to replace this background with another image. Existing approaches approximate α directly but post-processing is required to remove spill of the background colour in semi-transparent areas. Instead of estimating α directly, we propose 3 methods to estimate the unknown foreground colour, and then to deduce α. This approach leads to high quality mattes for transparent objects and allows spill-free results (see www.cs.nuim.ie/research/vision/data/imvip2014/). We show this through an evaluation of the proposed methods based on a ground truth dataset.]]>
Constant colour matting consists of estimating for each pixel of an image the propor- tion α of an unknown foreground colour with a known constant background colour. The α-matte is then used to replace this background with another image. Existing approaches approximate α directly but post-processing is required to remove spill of the background colour in semi-transparent areas. Instead of estimating α directly, we propose 3 methods to estimate the unknown foreground colour, and then to deduce α. This approach leads to high quality mattes for transparent objects and allows spill-free results (see www.cs.nuim.ie/research/vision/data/imvip2014/). We show this through an evaluation of the proposed methods based on a ground truth dataset.]]> Thu, 28 Aug 2014 12:44:36 GMT /slideshow/imvip2014/38461321 GuillaumeGales@slideshare.net(GuillaumeGales) Constant colour matting with foreground estimation GuillaumeGales Constant colour matting consists of estimating for each pixel of an image the propor- tion α of an unknown foreground colour with a known constant background colour. The α-matte is then used to replace this background with another image. Existing approaches approximate α directly but post-processing is required to remove spill of the background colour in semi-transparent areas. Instead of estimating α directly, we propose 3 methods to estimate the unknown foreground colour, and then to deduce α. This approach leads to high quality mattes for transparent objects and allows spill-free results (see www.cs.nuim.ie/research/vision/data/imvip2014/). We show this through an evaluation of the proposed methods based on a ground truth dataset. <img style="border:1px solid #C3E6D8;float:right;" alt="" src="https://cdn.slidesharecdn.com/ss_thumbnails/imvip2014-140828124436-phpapp02-thumbnail.jpg?width=120&height=120&fit=bounds" /> Constant colour matting consists of estimating for each pixel of an image the propor- tion α of an unknown foreground colour with a known constant background colour. The α-matte is then used to replace this background with another image. Existing approaches approximate α directly but post-processing is required to remove spill of the background colour in semi-transparent areas. Instead of estimating α directly, we propose 3 methods to estimate the unknown foreground colour, and then to deduce α. This approach leads to high quality mattes for transparent objects and allows spill-free results (see www.cs.nuim.ie/research/vision/data/imvip2014/). We show this through an evaluation of the proposed methods based on a ground truth dataset.

Constant colour matting with foreground estimation from Guillaume Gales

]]> 2341 3 https://cdn.slidesharecdn.com/ss_thumbnails/imvip2014-140828124436-phpapp02-thumbnail.jpg?width=120&height=120&fit=bounds presentation White http://activitystrea.ms/schema/1.0/post

http://activitystrea.ms/schema/1.0/posted

0 A Vision-Based Mobile Platform for Seamless Indoor/Outdoor Positioning /slideshow/a-visionbased-mobile-platform-for-seamless-indooroutdoor-positioning/17035555 weurogi2013-130308080612-phpapp02
The emergence of smartphones equipped with Internet access, high resolution cameras, and posi- tioning sensors opens up great opportunities for visualising geospatial information within augmented reality applications. While smartphones are able to provide geolocalisation, the inherent uncertainty in the estimated position, especially indoors, does not allow for completely accurate and robust alignment of the data with the camera images. In this paper we present a system that exploits computer vision techniques in conjunction with GPS and inertial sensors to create a seamless indoor/outdoor positioning vision-based platform. The vision-based approach estimates the pose of the camera relative to the fac ̧ade of a building and recognises the fac ̧ade from a georeferenced image database. This permits the insertion of 3D widgets into the user’s view with a known orientation relative to the fac ̧ade. For example, in Figure 1 (a) we show how this feature can be used to overlay directional information on the input image. Furthermore we provide an easy and intuitive interface for non-expert users to add their own georeferenced content to the system, encouraging volunteering GI. Indeed, to achieve this users only need to drag and drop predefined 3D widgets into a reference view of the fac ̧ade, see Figure 1 (b). The infrastructure is flexible in that we can add different layers of content on top of the fac ̧ades and hence, this opens many possibilities for different applications. Furthermore the system provides a representation suitable for both manual and automatic content authoring.]]>
The emergence of smartphones equipped with Internet access, high resolution cameras, and posi- tioning sensors opens up great opportunities for visualising geospatial information within augmented reality applications. While smartphones are able to provide geolocalisation, the inherent uncertainty in the estimated position, especially indoors, does not allow for completely accurate and robust alignment of the data with the camera images. In this paper we present a system that exploits computer vision techniques in conjunction with GPS and inertial sensors to create a seamless indoor/outdoor positioning vision-based platform. The vision-based approach estimates the pose of the camera relative to the fac ̧ade of a building and recognises the fac ̧ade from a georeferenced image database. This permits the insertion of 3D widgets into the user’s view with a known orientation relative to the fac ̧ade. For example, in Figure 1 (a) we show how this feature can be used to overlay directional information on the input image. Furthermore we provide an easy and intuitive interface for non-expert users to add their own georeferenced content to the system, encouraging volunteering GI. Indeed, to achieve this users only need to drag and drop predefined 3D widgets into a reference view of the fac ̧ade, see Figure 1 (b). The infrastructure is flexible in that we can add different layers of content on top of the fac ̧ades and hence, this opens many possibilities for different applications. Furthermore the system provides a representation suitable for both manual and automatic content authoring.]]> Fri, 08 Mar 2013 08:06:12 GMT /slideshow/a-visionbased-mobile-platform-for-seamless-indooroutdoor-positioning/17035555 GuillaumeGales@slideshare.net(GuillaumeGales) A Vision-Based Mobile Platform for Seamless Indoor/Outdoor Positioning GuillaumeGales The emergence of smartphones equipped with Internet access, high resolution cameras, and posi- tioning sensors opens up great opportunities for visualising geospatial information within augmented reality applications. While smartphones are able to provide geolocalisation, the inherent uncertainty in the estimated position, especially indoors, does not allow for completely accurate and robust alignment of the data with the camera images. In this paper we present a system that exploits computer vision techniques in conjunction with GPS and inertial sensors to create a seamless indoor/outdoor positioning vision-based platform. The vision-based approach estimates the pose of the camera relative to the fac ̧ade of a building and recognises the fac ̧ade from a georeferenced image database. This permits the insertion of 3D widgets into the user’s view with a known orientation relative to the fac ̧ade. For example, in Figure 1 (a) we show how this feature can be used to overlay directional information on the input image. Furthermore we provide an easy and intuitive interface for non-expert users to add their own georeferenced content to the system, encouraging volunteering GI. Indeed, to achieve this users only need to drag and drop predefined 3D widgets into a reference view of the fac ̧ade, see Figure 1 (b). The infrastructure is flexible in that we can add different layers of content on top of the fac ̧ades and hence, this opens many possibilities for different applications. Furthermore the system provides a representation suitable for both manual and automatic content authoring. <img style="border:1px solid #C3E6D8;float:right;" alt="" src="https://cdn.slidesharecdn.com/ss_thumbnails/weurogi2013-130308080612-phpapp02-thumbnail.jpg?width=120&height=120&fit=bounds" /> The emergence of smartphones equipped with Internet access, high resolution cameras, and posi- tioning sensors opens up great opportunities for visualising geospatial information within augmented reality applications. While smartphones are able to provide geolocalisation, the inherent uncertainty in the estimated position, especially indoors, does not allow for completely accurate and robust alignment of the data with the camera images. In this paper we present a system that exploits computer vision techniques in conjunction with GPS and inertial sensors to create a seamless indoor/outdoor positioning vision-based platform. The vision-based approach estimates the pose of the camera relative to the fac ̧ade of a building and recognises the fac ̧ade from a georeferenced image database. This permits the insertion of 3D widgets into the user’s view with a known orientation relative to the fac ̧ade. For example, in Figure 1 (a) we show how this feature can be used to overlay directional information on the input image. Furthermore we provide an easy and intuitive interface for non-expert users to add their own georeferenced content to the system, encouraging volunteering GI. Indeed, to achieve this users only need to drag and drop predefined 3D widgets into a reference view of the fac ̧ade, see Figure 1 (b). The infrastructure is flexible in that we can add different layers of content on top of the fac ̧ades and hence, this opens many possibilities for different applications. Furthermore the system provides a representation suitable for both manual and automatic content authoring.

A Vision-Based Mobile Platform for Seamless Indoor/Outdoor Positioning from Guillaume Gales

]]> 2261 4 https://cdn.slidesharecdn.com/ss_thumbnails/weurogi2013-130308080612-phpapp02-thumbnail.jpg?width=120&height=120&fit=bounds presentation Black http://activitystrea.ms/schema/1.0/post

http://activitystrea.ms/schema/1.0/posted

0 An Authoring Solution for a Façade-Based AR Platform: Infrastructure, Annotation and Visualization /slideshow/an-authoring-solution-for-a-faadebased-ar-platform-infrastructure-annotation-and-visualization/15021854 wismar2012-121104151337-phpapp02
�ݺ�ߣs from our presentation at ISMAR 2012 Workshop Authoring Solutions for Augmented Reality]]>
�ݺ�ߣs from our presentation at ISMAR 2012 Workshop Authoring Solutions for Augmented Reality]]> Sun, 04 Nov 2012 15:13:36 GMT /slideshow/an-authoring-solution-for-a-faadebased-ar-platform-infrastructure-annotation-and-visualization/15021854 GuillaumeGales@slideshare.net(GuillaumeGales) An Authoring Solution for a Façade-Based AR Platform: Infrastructure, Annotation and Visualization GuillaumeGales �ݺ�ߣs from our presentation at ISMAR 2012 Workshop Authoring Solutions for Augmented Reality <img style="border:1px solid #C3E6D8;float:right;" alt="" src="https://cdn.slidesharecdn.com/ss_thumbnails/wismar2012-121104151337-phpapp02-thumbnail.jpg?width=120&height=120&fit=bounds" /> �ݺ�ߣs from our presentation at ISMAR 2012 Workshop Authoring Solutions for Augmented Reality

An Authoring Solution for a Fa巽ade-Based AR Platform: Infrastructure, Annotation and Visualization from Guillaume Gales

]]> 1735 7 https://cdn.slidesharecdn.com/ss_thumbnails/wismar2012-121104151337-phpapp02-thumbnail.jpg?width=120&height=120&fit=bounds presentation Blanc http://activitystrea.ms/schema/1.0/post

http://activitystrea.ms/schema/1.0/posted

0 A Region-Based Randomized Voting Scheme for Stereo Matching /slideshow/a-regionbased-randomized-voting-scheme-for-stereo-matching/13949670 isvcslidesgales-120812081543-phpapp02
This paper presents a region-based stereo matching algo- rithm which uses a new method to select the nal disparity: a random process computes for each pixel dierent approximations of its disparity relying on a surface model with dierent image segmentations and each found disparity gets a vote. At last, the nal disparity is selected by estimating the mode of a density function built from these votes. We also advise how to choose the dierent parameters. Finally, an evaluation shows that the proposed method is ecient even at sub-pixel accuracy and is competitive with the state of the art.]]>
This paper presents a region-based stereo matching algo- rithm which uses a new method to select the nal disparity: a random process computes for each pixel dierent approximations of its disparity relying on a surface model with dierent image segmentations and each found disparity gets a vote. At last, the nal disparity is selected by estimating the mode of a density function built from these votes. We also advise how to choose the dierent parameters. Finally, an evaluation shows that the proposed method is ecient even at sub-pixel accuracy and is competitive with the state of the art.]]> Sun, 12 Aug 2012 08:15:40 GMT /slideshow/a-regionbased-randomized-voting-scheme-for-stereo-matching/13949670 GuillaumeGales@slideshare.net(GuillaumeGales) A Region-Based Randomized Voting Scheme for Stereo Matching GuillaumeGales This paper presents a region-based stereo matching algo- rithm which uses a new method to select the �nal disparity: a random process computes for each pixel di�erent approximations of its disparity relying on a surface model with di�erent image segmentations and each found disparity gets a vote. At last, the �nal disparity is selected by estimating the mode of a density function built from these votes. We also advise how to choose the di�erent parameters. Finally, an evaluation shows that the proposed method is e�cient even at sub-pixel accuracy and is competitive with the state of the art. <img style="border:1px solid #C3E6D8;float:right;" alt="" src="https://cdn.slidesharecdn.com/ss_thumbnails/isvcslidesgales-120812081543-phpapp02-thumbnail.jpg?width=120&height=120&fit=bounds" /> This paper presents a region-based stereo matching algo- rithm which uses a new method to select the �nal disparity: a random process computes for each pixel di�erent approximations of its disparity relying on a surface model with di�erent image segmentations and each found disparity gets a vote. At last, the �nal disparity is selected by estimating the mode of a density function built from these votes. We also advise how to choose the di�erent parameters. Finally, an evaluation shows that the proposed method is e�cient even at sub-pixel accuracy and is competitive with the state of the art.

A Region-Based Randomized Voting Scheme for Stereo Matching from Guillaume Gales

]]> 969 10 https://cdn.slidesharecdn.com/ss_thumbnails/isvcslidesgales-120812081543-phpapp02-thumbnail.jpg?width=120&height=120&fit=bounds presentation Black http://activitystrea.ms/schema/1.0/post

http://activitystrea.ms/schema/1.0/posted

0 Pixel matching for binocular stereovision by propagation of feature points matches and region-based randomized voting scheme /slideshow/soutenance7-en2/13949640 soutenance7en2-120812080141-phpapp01
Stereo matching is one of the main topics in computer vision. It consists in finding in two images of a same scene, taken from different viewpoints, the pairs of pixels which are the projections of a same scene point. Since the last twenty years, many local and global methods have been proposed to solve this problem. More recently, according to a reference evaluation protocol in the community, region-based methods showed interesting result in small-baseline binocular stereo (where images are taken nearby). The idea is to apply a colour segmentation algorithm on the images assuming that each pixel within a segment belongs to a same object surface. Then, the parameters of a surface model are computed, in the disparity space, for each segment according to initial disparities usually computed with a local method. Finally, a global optimization is performed to refine the results. A contribution of this thesis deals with a special kind of local method called seeds propagation. The search area of a correspondent is reduced to the neighbourhoods of reliable matches called seeds. This can help to reduce the computation time and to avoid some ambiguities. However, the success of such a method depends on the choice of these seeds. In this dissertation, we give a study of the seeds selection step. We focus on feature points matching. These are special points in the image with interesting characteristics for a given application. In our case, we need pixels that can be matched with high confidence. We compare fourteen different well-known detectors linked to five correlation measures. Some of these measures are meant to be robust to one of the main challenge in stereo matching: depth discontinuities. Besides, this study gives advice on the choice of the parameters of the different techniques to be able to find the best solutions according some given criteria. These parameters are estimated using machine learning. Then, these seeds are used with two approaches of propagation and the results are evaluated. Another contribution deals with a new region-based approach for dense stereo matching. Different colour segmentations are used. Then, many instances of a surface model are computed for the different regions according to initial disparities selected randomly. For each pixel, each instance gives a disparity value regarded as a vote. Finally, the most voted value is selected as the final disparity. This approach is relatively easy to implement and very effective giving competitive results among the state of the art.]]>
Stereo matching is one of the main topics in computer vision. It consists in finding in two images of a same scene, taken from different viewpoints, the pairs of pixels which are the projections of a same scene point. Since the last twenty years, many local and global methods have been proposed to solve this problem. More recently, according to a reference evaluation protocol in the community, region-based methods showed interesting result in small-baseline binocular stereo (where images are taken nearby). The idea is to apply a colour segmentation algorithm on the images assuming that each pixel within a segment belongs to a same object surface. Then, the parameters of a surface model are computed, in the disparity space, for each segment according to initial disparities usually computed with a local method. Finally, a global optimization is performed to refine the results. A contribution of this thesis deals with a special kind of local method called seeds propagation. The search area of a correspondent is reduced to the neighbourhoods of reliable matches called seeds. This can help to reduce the computation time and to avoid some ambiguities. However, the success of such a method depends on the choice of these seeds. In this dissertation, we give a study of the seeds selection step. We focus on feature points matching. These are special points in the image with interesting characteristics for a given application. In our case, we need pixels that can be matched with high confidence. We compare fourteen different well-known detectors linked to five correlation measures. Some of these measures are meant to be robust to one of the main challenge in stereo matching: depth discontinuities. Besides, this study gives advice on the choice of the parameters of the different techniques to be able to find the best solutions according some given criteria. These parameters are estimated using machine learning. Then, these seeds are used with two approaches of propagation and the results are evaluated. Another contribution deals with a new region-based approach for dense stereo matching. Different colour segmentations are used. Then, many instances of a surface model are computed for the different regions according to initial disparities selected randomly. For each pixel, each instance gives a disparity value regarded as a vote. Finally, the most voted value is selected as the final disparity. This approach is relatively easy to implement and very effective giving competitive results among the state of the art.]]> Sun, 12 Aug 2012 08:01:39 GMT /slideshow/soutenance7-en2/13949640 GuillaumeGales@slideshare.net(GuillaumeGales) Pixel matching for binocular stereovision by propagation of feature points matches and region-based randomized voting scheme GuillaumeGales Stereo matching is one of the main topics in computer vision. It consists in finding in two images of a same scene, taken from different viewpoints, the pairs of pixels which are the projections of a same scene point. Since the last twenty years, many local and global methods have been proposed to solve this problem. More recently, according to a reference evaluation protocol in the community, region-based methods showed interesting result in small-baseline binocular stereo (where images are taken nearby). The idea is to apply a colour segmentation algorithm on the images assuming that each pixel within a segment belongs to a same object surface. Then, the parameters of a surface model are computed, in the disparity space, for each segment according to initial disparities usually computed with a local method. Finally, a global optimization is performed to refine the results. A contribution of this thesis deals with a special kind of local method called seeds propagation. The search area of a correspondent is reduced to the neighbourhoods of reliable matches called seeds. This can help to reduce the computation time and to avoid some ambiguities. However, the success of such a method depends on the choice of these seeds. In this dissertation, we give a study of the seeds selection step. We focus on feature points matching. These are special points in the image with interesting characteristics for a given application. In our case, we need pixels that can be matched with high confidence. We compare fourteen different well-known detectors linked to five correlation measures. Some of these measures are meant to be robust to one of the main challenge in stereo matching: depth discontinuities. Besides, this study gives advice on the choice of the parameters of the different techniques to be able to find the best solutions according some given criteria. These parameters are estimated using machine learning. Then, these seeds are used with two approaches of propagation and the results are evaluated. Another contribution deals with a new region-based approach for dense stereo matching. Different colour segmentations are used. Then, many instances of a surface model are computed for the different regions according to initial disparities selected randomly. For each pixel, each instance gives a disparity value regarded as a vote. Finally, the most voted value is selected as the final disparity. This approach is relatively easy to implement and very effective giving competitive results among the state of the art. <img style="border:1px solid #C3E6D8;float:right;" alt="" src="https://cdn.slidesharecdn.com/ss_thumbnails/soutenance7en2-120812080141-phpapp01-thumbnail.jpg?width=120&height=120&fit=bounds" /> Stereo matching is one of the main topics in computer vision. It consists in finding in two images of a same scene, taken from different viewpoints, the pairs of pixels which are the projections of a same scene point. Since the last twenty years, many local and global methods have been proposed to solve this problem. More recently, according to a reference evaluation protocol in the community, region-based methods showed interesting result in small-baseline binocular stereo (where images are taken nearby). The idea is to apply a colour segmentation algorithm on the images assuming that each pixel within a segment belongs to a same object surface. Then, the parameters of a surface model are computed, in the disparity space, for each segment according to initial disparities usually computed with a local method. Finally, a global optimization is performed to refine the results. A contribution of this thesis deals with a special kind of local method called seeds propagation. The search area of a correspondent is reduced to the neighbourhoods of reliable matches called seeds. This can help to reduce the computation time and to avoid some ambiguities. However, the success of such a method depends on the choice of these seeds. In this dissertation, we give a study of the seeds selection step. We focus on feature points matching. These are special points in the image with interesting characteristics for a given application. In our case, we need pixels that can be matched with high confidence. We compare fourteen different well-known detectors linked to five correlation measures. Some of these measures are meant to be robust to one of the main challenge in stereo matching: depth discontinuities. Besides, this study gives advice on the choice of the parameters of the different techniques to be able to find the best solutions according some given criteria. These parameters are estimated using machine learning. Then, these seeds are used with two approaches of propagation and the results are evaluated. Another contribution deals with a new region-based approach for dense stereo matching. Different colour segmentations are used. Then, many instances of a surface model are computed for the different regions according to initial disparities selected randomly. For each pixel, each instance gives a disparity value regarded as a vote. Finally, the most voted value is selected as the final disparity. This approach is relatively easy to implement and very effective giving competitive results among the state of the art.

Pixel matching for binocular stereovision by propagation of feature points matches and region-based randomized voting scheme from Guillaume Gales

]]> 1214 20 https://cdn.slidesharecdn.com/ss_thumbnails/soutenance7en2-120812080141-phpapp01-thumbnail.jpg?width=120&height=120&fit=bounds presentation Black http://activitystrea.ms/schema/1.0/post

http://activitystrea.ms/schema/1.0/posted

0 Pixel Matching from Stereo Images (Callan seminar) /slideshow/pixel-matching-from-stereo-images-callan-seminar/13944534 callanseminar-120811103012-phpapp02
This talk discusses a number of techniques for correspondence estimation between stereo image pairs, i.e. two images of the same scene taken from different positions. The problem is to identify pairs of pixels in the two images that are the projections of the same scene point. Although the human visual system performs this task with ease, developing algorithms for automatically computing correspondences is a challenging task. In particular, existing algorithms can fail in homogeneous areas, near depth discontinuities and occlusions or with a repetitive texture pattern. The first part of this talk focuses on seed propagation-based approaches that are a special case of local methods based computing an iterative solution, where the solution is initialised using a sparse set of reliable matches (the seeds). I introduce a reliability measure used by the propagation technique for finding the correct correspondent of a pixel, providing robustness in the context of the above difficulties. This measure takes into account an unambiguity term, a continuity term and a colour consistency term. It has the advantage of taking into account information from the other candidates, and leads, according to our experimental evaluation, to better results when compared to other methods based on a correlation score alone. In the second part of this talk I will present ongoing work in our group on stereo matching in urban environments. In particular we exploit the fact that images of such environments contain multiple planar elements. I will show how utilising this strong geometrical constraint allows us to automatically segment building facades in single images. Furthermore I show how this technique permits robust pixel matching in wide-baseline stereo pairs. Finally, I will discuss how we intend to apply this technique for the development of augmented reality applications.]]>
This talk discusses a number of techniques for correspondence estimation between stereo image pairs, i.e. two images of the same scene taken from different positions. The problem is to identify pairs of pixels in the two images that are the projections of the same scene point. Although the human visual system performs this task with ease, developing algorithms for automatically computing correspondences is a challenging task. In particular, existing algorithms can fail in homogeneous areas, near depth discontinuities and occlusions or with a repetitive texture pattern. The first part of this talk focuses on seed propagation-based approaches that are a special case of local methods based computing an iterative solution, where the solution is initialised using a sparse set of reliable matches (the seeds). I introduce a reliability measure used by the propagation technique for finding the correct correspondent of a pixel, providing robustness in the context of the above difficulties. This measure takes into account an unambiguity term, a continuity term and a colour consistency term. It has the advantage of taking into account information from the other candidates, and leads, according to our experimental evaluation, to better results when compared to other methods based on a correlation score alone. In the second part of this talk I will present ongoing work in our group on stereo matching in urban environments. In particular we exploit the fact that images of such environments contain multiple planar elements. I will show how utilising this strong geometrical constraint allows us to automatically segment building facades in single images. Furthermore I show how this technique permits robust pixel matching in wide-baseline stereo pairs. Finally, I will discuss how we intend to apply this technique for the development of augmented reality applications.]]> Sat, 11 Aug 2012 10:30:10 GMT /slideshow/pixel-matching-from-stereo-images-callan-seminar/13944534 GuillaumeGales@slideshare.net(GuillaumeGales) Pixel Matching from Stereo Images (Callan seminar) GuillaumeGales This talk discusses a number of techniques for correspondence estimation between stereo image pairs, i.e. two images of the same scene taken from different positions. The problem is to identify pairs of pixels in the two images that are the projections of the same scene point. Although the human visual system performs this task with ease, developing algorithms for automatically computing correspondences is a challenging task. In particular, existing algorithms can fail in homogeneous areas, near depth discontinuities and occlusions or with a repetitive texture pattern. The first part of this talk focuses on seed propagation-based approaches that are a special case of local methods based computing an iterative solution, where the solution is initialised using a sparse set of reliable matches (the seeds). I introduce a reliability measure used by the propagation technique for finding the correct correspondent of a pixel, providing robustness in the context of the above difficulties. This measure takes into account an unambiguity term, a continuity term and a colour consistency term. It has the advantage of taking into account information from the other candidates, and leads, according to our experimental evaluation, to better results when compared to other methods based on a correlation score alone. In the second part of this talk I will present ongoing work in our group on stereo matching in urban environments. In particular we exploit the fact that images of such environments contain multiple planar elements. I will show how utilising this strong geometrical constraint allows us to automatically segment building facades in single images. Furthermore I show how this technique permits robust pixel matching in wide-baseline stereo pairs. Finally, I will discuss how we intend to apply this technique for the development of augmented reality applications. <img style="border:1px solid #C3E6D8;float:right;" alt="" src="https://cdn.slidesharecdn.com/ss_thumbnails/callanseminar-120811103012-phpapp02-thumbnail.jpg?width=120&height=120&fit=bounds" /> This talk discusses a number of techniques for correspondence estimation between stereo image pairs, i.e. two images of the same scene taken from different positions. The problem is to identify pairs of pixels in the two images that are the projections of the same scene point. Although the human visual system performs this task with ease, developing algorithms for automatically computing correspondences is a challenging task. In particular, existing algorithms can fail in homogeneous areas, near depth discontinuities and occlusions or with a repetitive texture pattern. The first part of this talk focuses on seed propagation-based approaches that are a special case of local methods based computing an iterative solution, where the solution is initialised using a sparse set of reliable matches (the seeds). I introduce a reliability measure used by the propagation technique for finding the correct correspondent of a pixel, providing robustness in the context of the above difficulties. This measure takes into account an unambiguity term, a continuity term and a colour consistency term. It has the advantage of taking into account information from the other candidates, and leads, according to our experimental evaluation, to better results when compared to other methods based on a correlation score alone. In the second part of this talk I will present ongoing work in our group on stereo matching in urban environments. In particular we exploit the fact that images of such environments contain multiple planar elements. I will show how utilising this strong geometrical constraint allows us to automatically segment building facades in single images. Furthermore I show how this technique permits robust pixel matching in wide-baseline stereo pairs. Finally, I will discuss how we intend to apply this technique for the development of augmented reality applications.

Pixel Matching from Stereo Images (Callan seminar) from Guillaume Gales

]]> 2344 8 https://cdn.slidesharecdn.com/ss_thumbnails/callanseminar-120811103012-phpapp02-thumbnail.jpg?width=120&height=120&fit=bounds presentation White http://activitystrea.ms/schema/1.0/post

http://activitystrea.ms/schema/1.0/posted

0 https://cdn.slidesharecdn.com/profile-photo-GuillaumeGales-48x48.jpg?cb=1713288101 Postdoc at NUIM Dept. of Computer Science https://cdn.slidesharecdn.com/ss_thumbnails/imvip2014-140828124436-phpapp02-thumbnail.jpg?width=320&height=320&fit=bounds slideshow/imvip2014/38461321 Constant colour mattin... https://cdn.slidesharecdn.com/ss_thumbnails/weurogi2013-130308080612-phpapp02-thumbnail.jpg?width=320&height=320&fit=bounds slideshow/a-visionbased-mobile-platform-for-seamless-indooroutdoor-positioning/17035555 A Vision-Based Mobile ... https://cdn.slidesharecdn.com/ss_thumbnails/wismar2012-121104151337-phpapp02-thumbnail.jpg?width=320&height=320&fit=bounds slideshow/an-authoring-solution-for-a-faadebased-ar-platform-infrastructure-annotation-and-visualization/15021854 An Authoring Solution ...