ABSTRACT Intelligent Robots and Computer Vision XV: Algorithms, Techniques, Active Vision, and Ma... more ABSTRACT Intelligent Robots and Computer Vision XV: Algorithms, Techniques, Active Vision, and Materials Handling, D.P. Casasent, Editor, Proc. SPIE 2904, 1996, pp. 239-249. One of the most important properties of neural networks is generality, as the same network can be trained to solve rather different tasks, depending on the training data. This is also one of the most prominent problems when practical real world problems are solved by neural networks, as existing domain knowledge is difficult to incorporate into the models. In this contribution we present methods for adding prior knowledge to neural network modeling. The approach is based on training the knowledge on the network instead of hard-coding the knowledge in advance to the connections or weights. The knowledge is specified as target values or constraints for different order partial derivatives of the network. This approach can be viewed as a flexible regularization method that controls directly the characteristics of the resulting ...
European Signal Processing Conference, Sep 1, 2000
In this paper, a new image compression method is presented using the Distance Transform on Curved... more In this paper, a new image compression method is presented using the Distance Transform on Curved Space (DTOCS) and derivative information in nding positions for control points. In previous work it has been shown that the control points are not in exactly optimal positions. This paper presents theoretical considerations according to which the new method enhances the decompressed image quality particularly in the areas of rapid changes. The obtained results shown verify the correctness of the theoretical considerations. The reconstructed image quality is clearly better measured by error criteria. Also visually the di erence is signi cant.
ABSTRACT The sequential mask operations for calculating distance transforms may have to be iterat... more ABSTRACT The sequential mask operations for calculating distance transforms may have to be iterated several times in the case of geodesic distances. This article presents an efficient propagation algorithm for the Distance Transform on Curved Space (DTOCS). It is based on a best-first pixel queue, and is applicable also for other gray-level distance transforms. It eliminates repetition of local distance calculations, and performs in near-linear time. A nearest neighbor transform based on distances along the surface, and a propagation direction image for tracing the shortest paths, can be produced simultaneously with the distance map.
This article presents an algorithm for finding and visualizing the shortest route between two poi... more This article presents an algorithm for finding and visualizing the shortest route between two points on a gray-level height map. The route is computed using gray-level distance transforms, which are variations of the Distance Transform on Curved Space (DTOCS). The basic Route DTOCS uses the chessboard kernel for calculating the distances between neighboring pixels, but variations, which take into account the larger distance between diagonal pixels, produce more accurate results, particularly for smooth and simple image surfaces. The route opimization algorithm is implemented using the Weighted Distance Transform on Curved Space (WDTOCS), which computes the piecewise Euclidean distance along the image surface, and the results are compared to the original Route DTOCS. The implementation of the algorithm is very simple, regardless of which distance definition is used.
Geodesic distance transforms are usually computed with sequential mask operations, which may have... more Geodesic distance transforms are usually computed with sequential mask operations, which may have to be iterated several times to get a globally optimal distance map. This article presents an efficient propagation algorithm based on a best-first pixel queue for computing the Distance Transform on Curved Space (DTOCS), applicable also for other geodesic distance transforms. It eliminates repetitions of local distance calculations, and performs in near-linear time.
This article presents an efficient priority pixel queue algorithm for calculating the distance tr... more This article presents an efficient priority pixel queue algorithm for calculating the distance transform on curved space (DTOCS), the corresponding nearest neighbor transform, and the new projection DTOCS (PDTOCS). The transforms provide tools for approximating distances and finding shortest paths on gray-level surfaces. Surface variation, or roughness, can also be measured.
The distance transform on curved space (DTOCS) and its locally Euclidean modification weighted DT... more The distance transform on curved space (DTOCS) and its locally Euclidean modification weighted DTOCS (WDTOCS) calculate distances along gray-level surfaces. This article presents the Route DTOCS algorithm for finding and visualizing the shortest route between two points on a gray-level height map, and also introduces new distance definitions producing more accurate global distances. The algorithm is very simple to implement,
The Distance Transform on Curved Space (DTOCS) can be used to calculate distances on a gray-level... more The Distance Transform on Curved Space (DTOCS) can be used to calculate distances on a gray-level surface, but the route along which the shortest distance is found, is lost during the calculation. In this article a new method for finding and visualizing the shortest path between two points on a gray-level height map is presented. The method is simple to implement, and example route images show that it produces good results.
In this paper we illustrate the use of a Code Camp approach in teaching programming skills. The C... more In this paper we illustrate the use of a Code Camp approach in teaching programming skills. The Code camp approach as an intensive and a social way of learning programming can be seen as a viable alternative to the traditional exercise based approach. Our experience is based on two separate implementations of the code camp method, a 24h and a one week long experiment. Approach is analyzed through extensive questionnaire. Based on the results of the questionnaire, the majority of students see the code camp a better approach than the traditional method. Therefore, it seems highly probable that the Code camp approach will be used extensively in the future for teaching programming skills.
One of the most important properties of neural networks is generality, as the same network can be... more One of the most important properties of neural networks is generality, as the same network can be trained to solve rather different tasks, depending on the training data. This is also one of the most prominent problems when practical real world problems are solved by neural networks, as existing domain knowledge is difficult to incorporate into the models. In this
Proceedings of the 12th International Conference on Discrete Geometry For Computer Imagery, 2005
Geodesic distance transforms are usually computed with sequential mask operations, which may have... more Geodesic distance transforms are usually computed with sequential mask operations, which may have to be iterated several times to get a globally optimal distance map. This article presents an efficient propagation algorithm based on a best-first pixel queue for computing the Distance Transform on Curved Space (DTOCS), applicable also for other geodesic distance transforms. It eliminates repetitions of local distance calculations, and performs in near-linear time.
Geodesic distance transforms are usually computed with sequential mask operations, which may have... more Geodesic distance transforms are usually computed with sequential mask operations, which may have to be iterated several times to get a globally optimal distance map. This article presents an efficient propagation algorithm based on a best-first pixel queue for computing the Distance Transform on Curved Space (DTOCS), applicable also for other geodesic distance transforms. It eliminates repetitions of local distance calculations, and performs in near-linear time.
ABSTRACT The sequential mask operations for calculating distance transforms may have to be iterat... more ABSTRACT The sequential mask operations for calculating distance transforms may have to be iterated several times in the case of geodesic distances. This article presents an efficient propagation algorithm for the Distance Transform on Curved Space (DTOCS). It is based on a best-first pixel queue, and is applicable also for other gray-level distance transforms. It eliminates repetition of local distance calculations, and performs in near-linear time. A nearest neighbor transform based on distances along the surface, and a propagation direction image for tracing the shortest paths, can be produced simultaneously with the distance map.
ABSTRACT Intelligent Robots and Computer Vision XV: Algorithms, Techniques, Active Vision, and Ma... more ABSTRACT Intelligent Robots and Computer Vision XV: Algorithms, Techniques, Active Vision, and Materials Handling, D.P. Casasent, Editor, Proc. SPIE 2904, 1996, pp. 239-249. One of the most important properties of neural networks is generality, as the same network can be trained to solve rather different tasks, depending on the training data. This is also one of the most prominent problems when practical real world problems are solved by neural networks, as existing domain knowledge is difficult to incorporate into the models. In this contribution we present methods for adding prior knowledge to neural network modeling. The approach is based on training the knowledge on the network instead of hard-coding the knowledge in advance to the connections or weights. The knowledge is specified as target values or constraints for different order partial derivatives of the network. This approach can be viewed as a flexible regularization method that controls directly the characteristics of the resulting ...
European Signal Processing Conference, Sep 1, 2000
In this paper, a new image compression method is presented using the Distance Transform on Curved... more In this paper, a new image compression method is presented using the Distance Transform on Curved Space (DTOCS) and derivative information in nding positions for control points. In previous work it has been shown that the control points are not in exactly optimal positions. This paper presents theoretical considerations according to which the new method enhances the decompressed image quality particularly in the areas of rapid changes. The obtained results shown verify the correctness of the theoretical considerations. The reconstructed image quality is clearly better measured by error criteria. Also visually the di erence is signi cant.
ABSTRACT The sequential mask operations for calculating distance transforms may have to be iterat... more ABSTRACT The sequential mask operations for calculating distance transforms may have to be iterated several times in the case of geodesic distances. This article presents an efficient propagation algorithm for the Distance Transform on Curved Space (DTOCS). It is based on a best-first pixel queue, and is applicable also for other gray-level distance transforms. It eliminates repetition of local distance calculations, and performs in near-linear time. A nearest neighbor transform based on distances along the surface, and a propagation direction image for tracing the shortest paths, can be produced simultaneously with the distance map.
This article presents an algorithm for finding and visualizing the shortest route between two poi... more This article presents an algorithm for finding and visualizing the shortest route between two points on a gray-level height map. The route is computed using gray-level distance transforms, which are variations of the Distance Transform on Curved Space (DTOCS). The basic Route DTOCS uses the chessboard kernel for calculating the distances between neighboring pixels, but variations, which take into account the larger distance between diagonal pixels, produce more accurate results, particularly for smooth and simple image surfaces. The route opimization algorithm is implemented using the Weighted Distance Transform on Curved Space (WDTOCS), which computes the piecewise Euclidean distance along the image surface, and the results are compared to the original Route DTOCS. The implementation of the algorithm is very simple, regardless of which distance definition is used.
Geodesic distance transforms are usually computed with sequential mask operations, which may have... more Geodesic distance transforms are usually computed with sequential mask operations, which may have to be iterated several times to get a globally optimal distance map. This article presents an efficient propagation algorithm based on a best-first pixel queue for computing the Distance Transform on Curved Space (DTOCS), applicable also for other geodesic distance transforms. It eliminates repetitions of local distance calculations, and performs in near-linear time.
This article presents an efficient priority pixel queue algorithm for calculating the distance tr... more This article presents an efficient priority pixel queue algorithm for calculating the distance transform on curved space (DTOCS), the corresponding nearest neighbor transform, and the new projection DTOCS (PDTOCS). The transforms provide tools for approximating distances and finding shortest paths on gray-level surfaces. Surface variation, or roughness, can also be measured.
The distance transform on curved space (DTOCS) and its locally Euclidean modification weighted DT... more The distance transform on curved space (DTOCS) and its locally Euclidean modification weighted DTOCS (WDTOCS) calculate distances along gray-level surfaces. This article presents the Route DTOCS algorithm for finding and visualizing the shortest route between two points on a gray-level height map, and also introduces new distance definitions producing more accurate global distances. The algorithm is very simple to implement,
The Distance Transform on Curved Space (DTOCS) can be used to calculate distances on a gray-level... more The Distance Transform on Curved Space (DTOCS) can be used to calculate distances on a gray-level surface, but the route along which the shortest distance is found, is lost during the calculation. In this article a new method for finding and visualizing the shortest path between two points on a gray-level height map is presented. The method is simple to implement, and example route images show that it produces good results.
In this paper we illustrate the use of a Code Camp approach in teaching programming skills. The C... more In this paper we illustrate the use of a Code Camp approach in teaching programming skills. The Code camp approach as an intensive and a social way of learning programming can be seen as a viable alternative to the traditional exercise based approach. Our experience is based on two separate implementations of the code camp method, a 24h and a one week long experiment. Approach is analyzed through extensive questionnaire. Based on the results of the questionnaire, the majority of students see the code camp a better approach than the traditional method. Therefore, it seems highly probable that the Code camp approach will be used extensively in the future for teaching programming skills.
One of the most important properties of neural networks is generality, as the same network can be... more One of the most important properties of neural networks is generality, as the same network can be trained to solve rather different tasks, depending on the training data. This is also one of the most prominent problems when practical real world problems are solved by neural networks, as existing domain knowledge is difficult to incorporate into the models. In this
Proceedings of the 12th International Conference on Discrete Geometry For Computer Imagery, 2005
Geodesic distance transforms are usually computed with sequential mask operations, which may have... more Geodesic distance transforms are usually computed with sequential mask operations, which may have to be iterated several times to get a globally optimal distance map. This article presents an efficient propagation algorithm based on a best-first pixel queue for computing the Distance Transform on Curved Space (DTOCS), applicable also for other geodesic distance transforms. It eliminates repetitions of local distance calculations, and performs in near-linear time.
Geodesic distance transforms are usually computed with sequential mask operations, which may have... more Geodesic distance transforms are usually computed with sequential mask operations, which may have to be iterated several times to get a globally optimal distance map. This article presents an efficient propagation algorithm based on a best-first pixel queue for computing the Distance Transform on Curved Space (DTOCS), applicable also for other geodesic distance transforms. It eliminates repetitions of local distance calculations, and performs in near-linear time.
ABSTRACT The sequential mask operations for calculating distance transforms may have to be iterat... more ABSTRACT The sequential mask operations for calculating distance transforms may have to be iterated several times in the case of geodesic distances. This article presents an efficient propagation algorithm for the Distance Transform on Curved Space (DTOCS). It is based on a best-first pixel queue, and is applicable also for other gray-level distance transforms. It eliminates repetition of local distance calculations, and performs in near-linear time. A nearest neighbor transform based on distances along the surface, and a propagation direction image for tracing the shortest paths, can be produced simultaneously with the distance map.
Uploads
Papers by Leena Ikonen