We developped a new interest point detector based on gradient orientations convergence. The aim is to reach better visual saliency than current detectors, such as Harris, Difference of gaussians or Loupias points. 

%center% %width=260% http://www-rocq.inria.fr/~ajoly/images/ http://www-rocq.inria.fr/~ajoly/images/
%center% '''left: 20 most salient points - right: 20 most rare points'''

! Constant tangential angle interest points [[http://www-rocq.inria.fr/~ajoly/MIR06.pdf| [MIR06] ]]

! Constant tangential angle interest points [[http://www-rocq.inria.fr/~ajoly/ajolyMIR05vf.pdf| [MIR06] ]]

! Density-based selection of local features [[http://www-rocq.inria.fr/~ajoly/ajolyMIR05vf.pdf| [MIR06] ]]

! Density-based selection of local features

! Content-based copy detection [[http://www-rocq.inria.fr/~ajoly/ajolyTMA07.pdf| [MIR07] ]]

! Content-based copy detection [[http://www-rocq.inria.fr/~ajoly/ajolyTMA06.pdf| [MIR07] ]] show-cases

! Content-based copy detection [[http://www-rocq.inria.fr/~ajoly/ajolyTMA06.pdf| [MIR06] ]] show-cases

! Content-based copy detection show-cases

! Constant tangential angle interest points
We developped a new interest point detector based on gradient orientations convergence. The aim is to reach better visual saliency than current detectors, such as Harris, Difference of gaussians or Loupias points. [[http://www-rocq.inria.fr/~ajoly/ajolyMIR05vf.pdf| [MIR06] ]]

We developped a new interest point detector based on gradient orientations convergence. The aim is to reach better visual saliency than current detectors, such as Harris, Difference of gaussians or Loupias points. [[http://www-rocq.inria.fr/~ajoly/ajolyMIR05vf.pdf| [1] ]]

[[#jolymir05]][[http://www-rocq.inria.fr/~ajoly/ajolyMIR05vf.pdf| [5] ]] "Discriminant Local Features Selection using Efficient Density Estimation in a Large Database", A. Joly and O. Buisson, ACM Int. Workshop on Multimedia Information Retrieval, invited paper, 2005.

There is an increasing variety of content-based image retrieval scenarios involving the use of local descriptors (object class recognition [[#Dork03| [1] ]], object and scene recognition [[#low04| [2] ]], content-based copy detection [[#joly05| [3] ]]). Enhancing the performance of these scenarios by using the geometric distribution or the relative positions of the local descriptors is an active research area. During this year, we have shown that in the copy detection scenario, the robust estimation of a global geometric transformation model after the search is widely profitable to improve the discrimination of the detection (paper submitted to IEEE transactions on multimedia). However, for other scenarios, using the geometry remains a challenging task: Including the geometric distribution in the descriptor itself often leads to a lake of robustness during the search of similar local descriptors whereas post-processing techniques are generally highly time consuming and thus limited to very small data sets. Moreover, in most of them, the geometric consistency is limited to rigid transformation models which do not allow to enforce the matching when two geometric distributions are dependent but not linearely linked. For a few months, we are investigating the use of non parametric geometric consistency measurements such as mutual information and robust correlation ratio and we plane to combine them with some robust local geometric properties that could be included in the descriptor itself in order to limit the number of matches during the second step.

This work was done in collaboration with the NII (National Institute of Japan) within the scope of my visit in Tokyo (july 2005).
Local features are well-suited to content-based image retrieval because of their locality, their local uniqueness and their high information content [[#Miko05 | [4] ]]. However, as they are selected only according to the local information content in the image, there is no guaranty that they will be distinctive in a large set of images. A local feature corresponding to a high saliency in the image can be highly redundant in some specific databases, such as the TV news database stored at NII in which textual characters are extremely frequent. To overcome this issue, we propose [[#jolymir05| [5] ]] to select relevant local features directly according to their discrimination power in a specific set of images. By computing the density of the local features in a source database with a new fast non parametric density estimation technique, it is indeed possible to select quickly the most ''rare'' local features in a large set of images. Figure illustrates the difference between the 20 most salient points of an image and the 20 most rare points according to their density in a large image database.

This work was done in collaboration with the NII (National Institute of Japan) within the scope of the visit of Alexis Joly at the NII (july 2005).

! Discriminant local features selection

%center% %width=260% http://www-rocq.inria.fr/~ajoly/images/femme_harris648.jpg http://www-rocq.inria.fr/~ajoly/images/femme_rares648.jpg | left: 20 most salient points - right: 20 most rare points

%center% %width=260% http://www-rocq.inria.fr/~ajoly/images/femme_harris648.jpg http://www-rocq.inria.fr/~ajoly/images/femme_rares648.jpg | %center% left: 20 most salient points - right: 20 most rare points

%center% %width=260% http://www-rocq.inria.fr/~ajoly/images/femme_harris648.jpg http://www-rocq.inria.fr/~ajoly/images/femme_rares648.jpg

%width=280% http://www-rocq.inria.fr/~ajoly/images/femme_harris648.jpg http://www-rocq.inria.fr/~ajoly/images/femme_rares648.jpg

%width=290% http://www-rocq.inria.fr/~ajoly/images/femme_harris648.jpg http://www-rocq.inria.fr/~ajoly/images/femme_rares648.jpg

%width=270% http://www-rocq.inria.fr/~ajoly/images/femme_harris648.jpg http://www-rocq.inria.fr/~ajoly/images/femme_rares648.jpg

%width=250% http://www-rocq.inria.fr/~ajoly/images/femme_harris648.jpg http://www-rocq.inria.fr/~ajoly/images/femme_rares648.jpg

%width=230% http://www-rocq.inria.fr/~ajoly/images/femme_harris648.jpg http://www-rocq.inria.fr/~ajoly/images/femme_rares648.jpg

%width=200% http://www-rocq.inria.fr/~ajoly/images/femme_harris648.jpg
%width=200% http://www-rocq.inria.fr/~ajoly/images/femme_rares648.jpg

%width=300% http://www-rocq.inria.fr/~ajoly/images/femme_rares648.jpg

%width=100% http://www-rocq.inria.fr/~ajoly/images/femme_rares648.jpg

http://www-rocq.inria.fr/~ajoly/images/femme_rares648.jpg

Local features are well-suited to content-based image retrieval because of their locality, their local uniqueness and their high information content [[#Miko05 | [4] ]]. However, as they are selected only according to the local information content in the image, there is no guaranty that they will be distinctive in a large set of images. A local feature corresponding to a high saliency in the image can be highly redundant in some specific databases, such as the TV news database stored at NII in which textual characters are extremely frequent. To overcome this issue, we propose [[#jolymir05| [5] ]] to select relevant local features directly according to their discrimination power in a specific set of images. By computing the density of the local features in a source database with a new fast non parametric density estimation technique, it is indeed possible to select quickly the most {\it rare} local features in a large set of images. Figure illustrates the difference between the 20 most salient points of an image and the 20 most rare points according to their density in a large image database.

Under Construction!!

[[#jolymir05]][[http://www-rocq.inria.fr/~ajoly/ajolyICIP05.pdf| [5] ]]

Local features are well-suited to content-based image retrieval because of their locality, their local uniqueness and their high information content [[#Miko05 | [4] ]]. However, as they are selected only according to the local information content in the image, there is no guaranty that they will be distinctive in a large set of images. A local feature corresponding to a high saliency in the image can be highly redundant in some specific databases, such as the TV news database stored at NII in which textual characters are extremely frequent. To overcome this issue, we propose \cite{joly05mir} to select relevant local features directly according to their discrimination power in a specific set of images. By computing the density of the local features in a source database with a new fast non parametric density estimation technique, it is indeed possible to select quickly the most {\it rare} local features in a large set of images. Figure illustrates the difference between the 20 most salient points of an image and the 20 most rare points according to their density in a large image database.

Local features are well-suited to content-based image retrieval because of their locality, their local uniqueness and their high information content \cite{Miko05Evaluation}. However, as they are selected only according to the local information content in the image, there is no guaranty that they will be distinctive in a large set of images. A local feature corresponding to a high saliency in the image can be highly redundant in some specific databases, such as the TV news database stored at NII in which textual characters are extremely frequent. To overcome this issue, we propose \cite{joly05mir} to select relevant local features directly according to their discrimination power in a specific set of images. By computing the density of the local features in a source database with a new fast non parametric density estimation technique, it is indeed possible to select quickly the most {\it rare} local features in a large set of images. Figure illustrates the difference between the 20 most salient points of an image and the 20 most rare points according to their density in a large image database.

[[#joly05]][[http://www-rocq.inria.fr/~ajoly/ajolyICIP05.pdf| [3] ]] "Content-based video copy detection in large databases: A local fingerprints  statistical similarity search approach", A. Joly, C. Frélicot and O. Buisson, in Proceedings of the Int. Conf. on Image Processing, 2005.

[[#joly05]][3] "Content-based video copy detection in large databases: A local fingerprints  statistical similarity search approach", A. Joly, C. Frélicot and O. Buisson, in Proceedings of the Int. Conf. on Image Processing, 2005 [[http://www-rocq.inria.fr/~ajoly/ajolyICIP05.pdf|pdf]].

There is an increasing variety of content-based image retrieval scenarios involving the use of local descriptors (object class recognition [[#Dork03| [1] ]], object and scene recognition [[#low04| [2] ]], content-based copy detection \cite{joly05icip}). Enhancing the performance of these scenarios by using the geometric distribution or the relative positions of the local descriptors is an active research area. During this year, we have shown that in the copy detection scenario, the robust estimation of a global geometric transformation model after the search is widely profitable to improve the discrimination of the detection (paper submitted to IEEE transactions on multimedia). However, for other scenarios, using the geometry remains a challenging task: Including the geometric distribution in the descriptor itself often leads to a lake of robustness during the search of similar local descriptors whereas post-processing techniques are generally highly time consuming and thus limited to very small data sets. Moreover, in most of them, the geometric consistency is limited to rigid transformation models which do not allow to enforce the matching when two geometric distributions are dependent but not linearely linked. For a few months, we are investigating the use of non parametric geometric consistency measurements such as mutual information and robust correlation ratio and we plane to combine them with some robust local geometric properties that could be included in the descriptor itself in order to limit the number of matches during the second step.

[[#low04]][2] "Distinctive image features from scale-invariant keypoints", D. Lowe, Int. Journal of Computer Vision, vol. 60, no. 2, pp. 91--110, 2004.

[[#Dork03]][1] "Selection of Scale-Invariant Parts for Object Class Recognition", G. Dorko, C. Schmid, IEEE Int. Conf. on Computer Vision, vol. 1, pp. 634--640, 2003.

There is an increasing variety of content-based image retrieval scenarios involving the use of local descriptors (object class recognition \cite{DS03}, object and scene recognition [[#low04| [1] ]], content-based copy detection \cite{joly05icip}). Enhancing the performance of these scenarios by using the geometric distribution or the relative positions of the local descriptors is an active research area. During this year, we have shown that in the copy detection scenario, the robust estimation of a global geometric transformation model after the search is widely profitable to improve the discrimination of the detection (paper submitted to IEEE transactions on multimedia). However, for other scenarios, using the geometry remains a challenging task: Including the geometric distribution in the descriptor itself often leads to a lake of robustness during the search of similar local descriptors whereas post-processing techniques are generally highly time consuming and thus limited to very small data sets. Moreover, in most of them, the geometric consistency is limited to rigid transformation models which do not allow to enforce the matching when two geometric distributions are dependent but not linearely linked. For a few months, we are investigating the use of non parametric geometric consistency measurements such as mutual information and robust correlation ratio and we plane to combine them with some robust local geometric properties that could be included in the descriptor itself in order to limit the number of matches during the second step.

[[#low04]][1] "Distinctive image features from scale-invariant keypoints", D. Lowe, Int. J. Computer Vision, vol. 60, no. 2, pp. 91--110, 2004.

There is an increasing variety of content-based image retrieval scenarios involving the use of local descriptors (object class recognition \cite{DS03}, object and scene recognition [[#low04|[1]]], content-based copy detection \cite{joly05icip}). Enhancing the performance of these scenarios by using the geometric distribution or the relative positions of the local descriptors is an active research area. During this year, we have shown that in the copy detection scenario, the robust estimation of a global geometric transformation model after the search is widely profitable to improve the discrimination of the detection (paper submitted to IEEE transactions on multimedia). However, for other scenarios, using the geometry remains a challenging task: Including the geometric distribution in the descriptor itself often leads to a lake of robustness during the search of similar local descriptors whereas post-processing techniques are generally highly time consuming and thus limited to very small data sets. Moreover, in most of them, the geometric consistency is limited to rigid transformation models which do not allow to enforce the matching when two geometric distributions are dependent but not linearely linked. For a few months, we are investigating the use of non parametric geometric consistency measurements such as mutual information and robust correlation ratio and we plane to combine them with some robust local geometric properties that could be included in the descriptor itself in order to limit the number of matches during the second step.

There is an increasing variety of content-based image retrieval scenarios involving the use of local descriptors (object class recognition \cite{DS03}, object and scene recognition [[#low04|1]], content-based copy detection \cite{joly05icip}). Enhancing the performance of these scenarios by using the geometric distribution or the relative positions of the local descriptors is an active research area. During this year, we have shown that in the copy detection scenario, the robust estimation of a global geometric transformation model after the search is widely profitable to improve the discrimination of the detection (paper submitted to IEEE transactions on multimedia). However, for other scenarios, using the geometry remains a challenging task: Including the geometric distribution in the descriptor itself often leads to a lake of robustness during the search of similar local descriptors whereas post-processing techniques are generally highly time consuming and thus limited to very small data sets. Moreover, in most of them, the geometric consistency is limited to rigid transformation models which do not allow to enforce the matching when two geometric distributions are dependent but not linearely linked. For a few months, we are investigating the use of non parametric geometric consistency measurements such as mutual information and robust correlation ratio and we plane to combine them with some robust local geometric properties that could be included in the descriptor itself in order to limit the number of matches during the second step.

[[#low04]] "Distinctive image features from scale-invariant keypoints", D. Lowe, Int. J. Computer Vision, vol. 60, no. 2, pp. 91--110, 2004.

There is an increasing variety of content-based image retrieval scenarios involving the use of local descriptors (object class recognition \cite{DS03}, object and scene recognition \cite{Lowe04}, content-based copy detection \cite{joly05icip}). Enhancing the performance of these scenarios by using the geometric distribution or the relative positions of the local descriptors is an active research area. During this year, we have shown that in the copy detection scenario, the robust estimation of a global geometric transformation model after the search is widely profitable to improve the discrimination of the detection (paper submitted to IEEE transactions on multimedia). However, for other scenarios, using the geometry remains a challenging task: Including the geometric distribution in the descriptor itself often leads to a lake of robustness during the search of similar local descriptors whereas post-processing techniques are generally highly time consuming and thus limited to very small data sets. Moreover, in most of them, the geometric consistency is limited to rigid transformation models which do not allow to enforce the matching when two geometric distributions are dependent but not linearely linked. For a few months, we are investigating the use of non parametric geometric consistency measurements such as mutual information and robust correlation ratio and we plane to combine them with some robust local geometric properties that could be included in the descriptor itself in order to limit the number of matches during the second step.

Local features are well-suited to content-based image retrieval because of their locality, their local uniqueness and their high information content \cite{Miko05Evaluation}. However, as they are selected only according to the local information content in the image, there is no guaranty that they will be distinctive in a large set of images. A local feature corresponding to a high saliency in the image can be highly redundant in some specific databases, such as the TV news database stored at NII in which textual characters are extremely frequent. To overcome this issue, we propose \cite{joly05mir} to select relevant local features directly according to their discrimination power in a specific set of images. By computing the density of the local features in a source database with a new fast non parametric density estimation technique, it is indeed possible to select quickly the most {\it rare} local features in a large set of images. Figure illustrates the difference between the 20 most salient points of an image and the 20 most rare points according to their density in a large image database.

\subsection{Discriminant local features selection}
Keywords: image retrieval, local features, discriminant, density estimation 
Participants: Alexis Joly, Nozha Boujemaa, Shin'ichi Satoh\\

Under Construction!!

In progress !!