[attachment=4973]
Scope of the project:
The aim is to build a fast and efficient strategy to retrieve the query
Concept in content-based image retrieval
Introduction:
1) RGB Projections:
The RGB color model is an additive color model in which red, green, and blue light are added together in various ways to reproduce a broad array of colors. The name of the model comes from the initials of the three additive primary colors, red, green, and blue. The main purpose of the RGB color model is for the sensing, representation, and display of images in electronic systems, such as conventional photography.
In this module the RGB Projections is used to find the size of the image vertically and horizontally.
2) Image Utility:
Whenever minimizing the error of classification is interesting for CBIR, this criterion does not completely reflect the user satisfaction. Other utility criteria
Closer to this, such as precision, should provide more efficient selections.
3) Comparable Image:
In this module a reselection technique to speed up the selection process, which leads to a computational complexity negligible compared to the size of the database for the whole active learning process. All these components are integrated in our retrieval system, called RETIN and the user gives new labels for images, and they are compared to the current classification. If the user mostly gives relevant labels, the system should propose new images for labeling around a higher rank to get more irrelevant labels.
4) Similarity measure:
The results in terms of mean average precision according to the training set size (we omit the KFD which gives results very close to inductive SVMs) for both ANN and Corel databases. One can see that the classification-based methods give the best results, showing the power of statistical methods over geometrical approaches, like the one reported here (similarity refinement method).
5) Result:
Finally, the image will take the relevant image what the user search. One can see that we have selected concepts of different levels of complexities. The performances go from few percentages of Mean average precision to 89%. The concepts that are the most difficult to retrieve are very small and/or have a much diversified visual content. The method which aims at minimizing the error of generalization is the less efficient active learning method. The most efficient method is the precision- oriented method.
Active Learning Methods for Interactive Image Retrieval
Scope of the project:
The aim is to build a fast and efficient strategy to retrieve the query
Concept in content-based image retrieval
Introduction:
Human interactive systems have attracted a lot of research interest in recent years, especially for content- based image retrieval systems. Contrary to the early systems, which focused on fully automatic strategies, recent approaches have introduced human-computer interaction. In this paper, we focus on the retrieval of concepts within a large image collection. We assume that a user is looking for a set of images, the query concept, within a database. The aim is to build a fast and efficient strategy to retrieve the query
Concept. In content-based image retrieval (CBIR), the search may be initiated using a query as an example. The top rank similar images are then presented to the user. Then, the interactive process allows the user to refine his request as much as necessary in
a relevance feedback loop. Many kinds of interaction between the user and the system have been proposed, but most of the time, user information consists of binary labels indicating whether or not the image belongs to the desired concept.
Module Description:Concept. In content-based image retrieval (CBIR), the search may be initiated using a query as an example. The top rank similar images are then presented to the user. Then, the interactive process allows the user to refine his request as much as necessary in
a relevance feedback loop. Many kinds of interaction between the user and the system have been proposed, but most of the time, user information consists of binary labels indicating whether or not the image belongs to the desired concept.
1) RGB Projections:
The RGB color model is an additive color model in which red, green, and blue light are added together in various ways to reproduce a broad array of colors. The name of the model comes from the initials of the three additive primary colors, red, green, and blue. The main purpose of the RGB color model is for the sensing, representation, and display of images in electronic systems, such as conventional photography.
In this module the RGB Projections is used to find the size of the image vertically and horizontally.
2) Image Utility:
Whenever minimizing the error of classification is interesting for CBIR, this criterion does not completely reflect the user satisfaction. Other utility criteria
Closer to this, such as precision, should provide more efficient selections.
3) Comparable Image:
In this module a reselection technique to speed up the selection process, which leads to a computational complexity negligible compared to the size of the database for the whole active learning process. All these components are integrated in our retrieval system, called RETIN and the user gives new labels for images, and they are compared to the current classification. If the user mostly gives relevant labels, the system should propose new images for labeling around a higher rank to get more irrelevant labels.
4) Similarity measure:
The results in terms of mean average precision according to the training set size (we omit the KFD which gives results very close to inductive SVMs) for both ANN and Corel databases. One can see that the classification-based methods give the best results, showing the power of statistical methods over geometrical approaches, like the one reported here (similarity refinement method).
5) Result:
Finally, the image will take the relevant image what the user search. One can see that we have selected concepts of different levels of complexities. The performances go from few percentages of Mean average precision to 89%. The concepts that are the most difficult to retrieve are very small and/or have a much diversified visual content. The method which aims at minimizing the error of generalization is the less efficient active learning method. The most efficient method is the precision- oriented method.
