Jari Kangas | Tampere University (original) (raw)
Papers by Jari Kangas
| Image compression is an essential task for image storage and transmission applications. Vector ... more | Image compression is an essential task for image storage and transmission applications. Vector quantization is often used when high compression rates are needed. Self-Organizing Map (SOM) algorithm can be used to generate codebooks for vector quantization. Previously it has been demonstrated that using the special property of the SOM algorithm that the codebook entries are ordered one can use prediction coding of codewords to make the compression more e ective. In this paper it is shown that training the SOM algorithm by using di erent weighting for sample blocks having di erent statistical characteristics one can further increase the compression e ciency.
Image compression is needed for image storage and transmission applications.
... Map method has also been used in Japanese tests for more thorough classification of nasals /m... more ... Map method has also been used in Japanese tests for more thorough classification of nasals /m,n,ri ... amount to 16 to 17 percent of all phonemes in Finnish.) The Transient Map method ... REFERENCES 1. KOHONEN, T. and REUHKALA, E., "A Very Fast Associative Method for the ...
Second International Conference on Image and Graphics, 2002
ABSTRACT In this paper, a normalized RGB space based color clustering method is given, and furthe... more ABSTRACT In this paper, a normalized RGB space based color clustering method is given, and further the application of color clutering in character location is described. Color clustering is used to group a color image into the different binary layers. During color clustering, the normalized RGB space is adopted. In each layer, its color should be homogeneous. As characters normally have different information in color from their background, characters and their background are grouped into different color layers, which are fairly useful to locate characters. In order to achieve character location, an aligning and merging analysis (AMA) scheme is presented to locate all potential characters on each color layer. The experimental results have proven the effectiveness of the method, which is one important part of the optical character recognition (OCR) system.
Artificial Neural Networks, 1991
This paper proposes a method of identifying character-like regions in order to extract and recogn... more This paper proposes a method of identifying character-like regions in order to extract and recognize characters in natural color scene images automatically. After connected component extraction based on a multi-group decomposition scheme, alignment analysis is used to check the block candidates, namely, the character-like regions in each binary image layer and the final composed image. Priority adaptive segmentation (PAS) is
This paper deals with the problem of finding good prototypes for a condensed nearest neighbor cla... more This paper deals with the problem of finding good prototypes for a condensed nearest neighbor classifier for a character recognition system based on directional codes. A prototype search by a genetic algorithm capable of creating new prototypes is compared against a direct prototype selection algorithm. It is shown in a leave-one-out experiment that the prototypes found by a genetic algorithm give significantly smaller intraclass distances to class members, and significantly larger normalized interclass distances.
... Map method has also been used in Japanese tests for more thorough classification of nasals /m... more ... Map method has also been used in Japanese tests for more thorough classification of nasals /m,n,ri ... amount to 16 to 17 percent of all phonemes in Finnish.) The Transient Map method ... REFERENCES 1. KOHONEN, T. and REUHKALA, E., "A Very Fast Associative Method for the ...
Image compression is needed for image storage and transmission applications.
Icann, 1991
In connection to a speech recognizer, the aim of which is to produce phonemic transcriptions of a... more In connection to a speech recognizer, the aim of which is to produce phonemic transcriptions of arbitrary spoken utterances, we investigate the combined e ect of several improvements at di erent stages of phoneme recognition. The core of the basic recognition system is Learning Vector Quantization (LVQ1) 1]. This algorithm was originally used to classify FFT-based short-time feature vectors into phonemic classes. The phonemic decoding stage was earlier based on simple durational rules 2] 3]. At the feature level, we now study the e ect of using mel-scale cepstral features and concatenating consecutive feature vectors to include context. At the output of vector quantization, a comparison of three approaches to take into account the classi cations of feature vectors in local context is presented. The rule-based phonemic decoding is compared to decoding employing Hidden Markov Models (HMMs). As earlier, an optional grammatical post-correction method (DEC) is applied. Experiments conducted with three male speakers indicate that it is possible to increase signi cantly the phonemic transcription accuracy of the previous conguration. By using appropriately liftered cepstra, concatenating three adjacent feature vectors, and using HMM-based phonemic decoding, the error rate can be decreased from 14.0 % to 5.8 %.
[Proceedings 1992] IJCNN International Joint Conference on Neural Networks, 1992
This paper is an overview of the program package LVQ PAK, which has been developed for convenient... more This paper is an overview of the program package LVQ PAK, which has been developed for convenient and e ective application of Learning Vector Quantization algorithms. Two new features are included: fast con ict-free initial distribution of codebook vectors into the class zones, and the optimized-learning-rate algorithm OLVQ1.
| Image compression is an essential task for image storage and transmission applications. Vector ... more | Image compression is an essential task for image storage and transmission applications. Vector quantization is often used when high compression rates are needed. Self-Organizing Map (SOM) algorithm can be used to generate codebooks for vector quantization. Previously it has been demonstrated that using the special property of the SOM algorithm that the codebook entries are ordered one can use prediction coding of codewords to make the compression more e ective. In this paper it is shown that training the SOM algorithm by using di erent weighting for sample blocks having di erent statistical characteristics one can further increase the compression e ciency.
Image compression is needed for image storage and transmission applications.
... Map method has also been used in Japanese tests for more thorough classification of nasals /m... more ... Map method has also been used in Japanese tests for more thorough classification of nasals /m,n,ri ... amount to 16 to 17 percent of all phonemes in Finnish.) The Transient Map method ... REFERENCES 1. KOHONEN, T. and REUHKALA, E., "A Very Fast Associative Method for the ...
Second International Conference on Image and Graphics, 2002
ABSTRACT In this paper, a normalized RGB space based color clustering method is given, and furthe... more ABSTRACT In this paper, a normalized RGB space based color clustering method is given, and further the application of color clutering in character location is described. Color clustering is used to group a color image into the different binary layers. During color clustering, the normalized RGB space is adopted. In each layer, its color should be homogeneous. As characters normally have different information in color from their background, characters and their background are grouped into different color layers, which are fairly useful to locate characters. In order to achieve character location, an aligning and merging analysis (AMA) scheme is presented to locate all potential characters on each color layer. The experimental results have proven the effectiveness of the method, which is one important part of the optical character recognition (OCR) system.
Artificial Neural Networks, 1991
This paper proposes a method of identifying character-like regions in order to extract and recogn... more This paper proposes a method of identifying character-like regions in order to extract and recognize characters in natural color scene images automatically. After connected component extraction based on a multi-group decomposition scheme, alignment analysis is used to check the block candidates, namely, the character-like regions in each binary image layer and the final composed image. Priority adaptive segmentation (PAS) is
This paper deals with the problem of finding good prototypes for a condensed nearest neighbor cla... more This paper deals with the problem of finding good prototypes for a condensed nearest neighbor classifier for a character recognition system based on directional codes. A prototype search by a genetic algorithm capable of creating new prototypes is compared against a direct prototype selection algorithm. It is shown in a leave-one-out experiment that the prototypes found by a genetic algorithm give significantly smaller intraclass distances to class members, and significantly larger normalized interclass distances.
... Map method has also been used in Japanese tests for more thorough classification of nasals /m... more ... Map method has also been used in Japanese tests for more thorough classification of nasals /m,n,ri ... amount to 16 to 17 percent of all phonemes in Finnish.) The Transient Map method ... REFERENCES 1. KOHONEN, T. and REUHKALA, E., "A Very Fast Associative Method for the ...
Image compression is needed for image storage and transmission applications.
Icann, 1991
In connection to a speech recognizer, the aim of which is to produce phonemic transcriptions of a... more In connection to a speech recognizer, the aim of which is to produce phonemic transcriptions of arbitrary spoken utterances, we investigate the combined e ect of several improvements at di erent stages of phoneme recognition. The core of the basic recognition system is Learning Vector Quantization (LVQ1) 1]. This algorithm was originally used to classify FFT-based short-time feature vectors into phonemic classes. The phonemic decoding stage was earlier based on simple durational rules 2] 3]. At the feature level, we now study the e ect of using mel-scale cepstral features and concatenating consecutive feature vectors to include context. At the output of vector quantization, a comparison of three approaches to take into account the classi cations of feature vectors in local context is presented. The rule-based phonemic decoding is compared to decoding employing Hidden Markov Models (HMMs). As earlier, an optional grammatical post-correction method (DEC) is applied. Experiments conducted with three male speakers indicate that it is possible to increase signi cantly the phonemic transcription accuracy of the previous conguration. By using appropriately liftered cepstra, concatenating three adjacent feature vectors, and using HMM-based phonemic decoding, the error rate can be decreased from 14.0 % to 5.8 %.
[Proceedings 1992] IJCNN International Joint Conference on Neural Networks, 1992
This paper is an overview of the program package LVQ PAK, which has been developed for convenient... more This paper is an overview of the program package LVQ PAK, which has been developed for convenient and e ective application of Learning Vector Quantization algorithms. Two new features are included: fast con ict-free initial distribution of codebook vectors into the class zones, and the optimized-learning-rate algorithm OLVQ1.