# Timo Kaukoranta's research while affiliated with University of Turku and other places

**What is this page?**

This page lists the scientific contributions of an author, who either does not have a ResearchGate profile, or has not yet added these contributions to their profile.

It was automatically created by ResearchGate to create a record of this author's body of work. We create such pages to advance our goal of creating and maintaining the most comprehensive scientific repository possible. In doing so, we process publicly available (personal) data relating to the author as a member of the scientific community.

If you're a ResearchGate member, you can follow this page to keep up with this author's work.

If you are this author, and you don't want us to display this page anymore, please let us know.

It was automatically created by ResearchGate to create a record of this author's body of work. We create such pages to advance our goal of creating and maintaining the most comprehensive scientific repository possible. In doing so, we process publicly available (personal) data relating to the author as a member of the scientific community.

If you're a ResearchGate member, you can follow this page to keep up with this author's work.

If you are this author, and you don't want us to display this page anymore, please let us know.

## Publications (48)

Clustering of a data set can be done by the well-known Pairwise Nearest Neighbor (PNN) algorithm. The algorithm is conceptionally very simple and gives high quality solutions. A drawback of the method is the relatively large running time of the original (exact) implementation. Recently, an efficient version of the exact PNN algorithm has been intro...

We propose a new iterative a lgorithm for the generation o f a c odebook in v ector quantization. The a lgorithm starts with an initial codebook that i s improved b y a combination o f merge a nd split operations. By merging small neighboring clusters, additional resources (codevectors) will be released. These e xtra c odevectors can b e reallocate...

Agglomerative clustering generates the partition hierarchically by a sequence of merge operations. We propose an alternative to the merge-based approach by removing the clusters iteratively one by one until the desired number of clusters is reached. We apply local optimization strategy by always removing the cluster that increases the distortion th...

Straightforward implementation of the exact pairwise nearest neighbor (PNN) algorithm takes ( 3 ) time, where is the number of training vectors. This is rather slow in practical situations. Fortunately, much faster implementation can be obtained with rather simple modifications to the basic algorithm. In this paper, we propose a fast ( 2 ) time imp...

We consider the clustering problem in the case where the distances between elements are metric and both the number of attributes and the number of clusters are large. In this environment the genetic algorithm approach gives high quality clusterings, but at the expense of long running time. Three new and efficient crossover techniques are introduced...

This paper introduces a new method for reducing the number of distance calculations in the generalized Lloyd algorithm (GLA), which is a widely used method to construct a codebook in vector quantization. Reduced comparison search detects the activity of the code vectors and utilizes it on the classification of the training vectors. For training vec...

We propose a new iterative algorithm for the generation of a codebook in vector quantization. The algorithm starts with an initial codebook that is improved by a combination of merge and split operations. By merging small neighboring clusters, additional resources (codevectors) will be released. These extra codevectors can be reallocated by splitti...

The pairwise nearest neighbor (PNN) method is a simple and well-known method for codebook generation in vector quantization. In its exact form, it provides a good-quality codebook but at the cost of high run time. A fast exact algorithm was recently introduced to implement the PNN an order of magnitude faster than the original O(N K) time algorithm...

Soft centroids method is proposed for binary vector quantizer design. Instead of using binary centroids, the codevectors can take any real value between one and zero during the codebook generation process. The binarization is performed only for the final codebook. The proposed method is successfully applied for three existing codebook generation al...

Scalar quantization is a basic technique for analog-to-digital signal transformation. It has an extensive theoretical background in addition to the practical usefulness. Vector quantization offers very good possibilities for lossy signal compression. The method has the great advance that its decoding runs extremely fast. Unfortunately, it includes...

We discuss pattern recognition in the context of computer game. The purpose of pattern recognition is to extract relevant information from the game world. This high level information is needed by a decision-making system, which is responsible for producing actions to the game world. We delineate where pattern recognition can be applied in computer...

We introduce a computer game platform, AIsHockey, which is based on the real-world game of ice hockey. The platform allows us to implement and study autonomous, real-time synthetic players (i.e., computer-controlled actors in a game). By applying the Model-View-Controller architectural pattern we define the role of a synthetic player and recognize...

As game worlds resemble the real world more closely, their increasing complexity requires effective and reliable pattern recognition. Computer games encompass a variety of problems involving pattern recognition including enemy evaluation and prediction, coaching, group coordination, terrain analysis, and learning. The task of pattern recognition is...

Multiplayer computer games rely on the practicability of the underlying network infrastructure. In this paper, we describe essential networking concepts and review techniques developed for reducing networking resource requirements in distributed interactive real-time applications. Also, we present a survey of the relevant literature concentrating o...

Distributed, real-time multiplayer computer games (MCGs) are in the vanguard of utilizing the networking possibilities. Although related research have been done in military simulations, virtual reality systems, and computer supported cooperative working, the suggested solutions diverge from the problems posed by MCGs. With this in mind, this paper...

The pairwise nearest neighbor method (PNN) generates the clustering of a given data set by a sequence of merge steps. We propose an alternative solution for the merge-based approach by introducing an iterative shrinking method. The new method removes the clusters iteratively one by one until the desired number of clusters is reached. Instead of mer...

Networking forms an essential part of multiplayer computer games. In this paper, we review the techniques developed for improving networking in distributed interactive real-time applications. We present a survey of the relevant literature concentrating on the research done on military simulations, networked virtual environments, and multiplayer com...

Networking forms an essential part of multiplayer computer games. In this paper, we review the techniques developed for improving networking in distributed interactive real-time applications. We present a survey of the relevant literature concentrating on the research done on military simulations, networked virtual environments, and multiplayer com...

The problem of finding optimal clustering has not been well covered in the literature. Solutions can be found only for special cases, which can be solved in polynomial time. In this paper, we give a solution for the general case. The method generates all possible clusterings by a series of merge steps. The clusterings are organized as a minimum red...

Distributed, real-time multiplayer computer games (MCGs) are in the vanguard of utilizing the networking possibilities. Although related research have been done in military simulations, virtual reality systems, and computer supported cooperative working, the suggested solutions diverge from the problems posed by MCGs. With this in mind, this paper...

The pairwise nearest neighbor (PNN) method is a simple and well-known method for codebook generation in vector quantization. In its exact form, it provides a good-quality codebook but at the cost of high run time. A fast exact algorithm was recently introduced to implement the PNN an order of magnitude faster than the original O(N3K) time algorithm...

Pairwise nearest neighbor method (PNN), in its exact form, provides good quality codebooks for vector quantization but at the cost of high run time. We propose mean-distance ordered search to reduce the amount of work caused by the distance calculations. The method was originally proposed for the encoding step of the vector quantization. We modify...

Pairwise nearest neighbor method (PNN), in its exact form, provides good quality codebooks for vector quantization but at
the cost of high run time. We consider the utilization of the partial distortion search technique in order to reduce the workload
caused by the distance calculations in the PNN. By experiments, we show that the simple improvemen...

Straightforward implementation of the exact pairwise nearest neighbor (PNN) algorithm takes O(N3) time, where N is the number of training vectors. This is rather slow in practical situations. Fortunately, much faster implementation can be obtained with rather simple modifications to the basic algorithm. In this paper, we propose a fast O(tauN2) tim...

This paper introduces a new method for reducing the number of distance calculations in the generalized Lloyd algorithm (GLA), which is a widely used method to construct a codebook in vector quantization. Reduced comparison search detects the activity of the code vectors and utilizes it on the classification of the training vectors. For training vec...

Straightforward implementation of the exact pairwise nearest neighbor (PNN) takes O(N 3 ) time, where N is the number of training vectors. This is rather slow in practical situations. Fortunately much faster implementation can be obtained with rather simple modifications to the basic algorithm. In the present paper we propose a fast O(tN 2 ) time i...

This paper introduces a new method for reducing the number of
distance calculations in the generalized Lloyd algorithm (GLA), which is
a widely used method to construct a codebook in vector quantization. The
reduced comparison search detects the activity of the code vectors and
utilizes it on the classification of the training vectors. For training...

Straightforward implementation of the exact pairwise nearest neighbor (PNN) takes O(N³) time, where N is the number of training vectors. This is rather slow in practical situations. Fortunately much faster implementation can be obtained with rather simple modifications to the basic algorithm. In the present paper we propose a fast O(tN²) time imple...

The pairwise nearest neighbor (PNN) algorithm is a well- known method for the codebook construction in vector quantization and for the clustering of data sets. The algorithm has a simple structure and it provides high quality solutions. A drawback of the method is the large running time of the original (exact) implementation. We prove the monotony...

We consider the codebook generation problem involved in the design
of a vector quantizer. The aim is to find M code vectors (codebook) for
a given set of N training vectors (training set) by minimizing the
average pairwise distance between the training vectors and their
representative code vectors. Straightforward implementation of the
optimal pair...

We propose a new iterative algorithm for the generation of a
codebook in vector quantization. The algorithm starts with an initial
codebook that is improved by a sequence of merge and split operations.
By merging small neighboring clusters additional resources (code
vectors) will be released. These extra code vectors can be reallocated
by splitting...

We propose a new iterative algorithm for the generation of a codebook in vector quantization. The algorithm starts with an initial codebook that is improved by a combination of merge and split operations. By merging small neighboring clusters, additional resources (codevectors) are released. These extra codevectors can be reallocated by splitting l...

In this work, lossless grayscale image compression methods are compared on a medical image database. The database contains 10 different types of images with bit rates varying from 8 to 16 bits per pixel. The total number of test images was about 3000, originating from 125 different patient studies. Methods used for compressing the images include se...

The well-known LBG algorithm uses binary splitting for generating an initial codebook, which is then iteratively improved by the generalized Lloyd algorithm (GLA). We study different variants of the splitting method and its application to codebook generation with and without the GLA. A new iterative splitting method is proposed, which is applicable...

Abstract The well known LBG algorithm uses binary splitting method for generating initial codebook which is then iteratively improved by GLA. In the present,paper we study different variants of the splitting method and its application,to codebook generation problem - with and without GLA. A new iterative splitt,ing method is proposed which is appli...

The performance of the generalised Lloyd algorithm (GLA) is
improved by reallocating the codevectors every time the GLA reaches a
local optimum. This is achieved by splitting the largest partition and
by merging two small neighbouring partitions, thereby preserving the
size of the codebook. The whole procedure is repeated until no
improvement is ac...

In the present work we study image quality evaluation from an empirical point of view and want to find out (1) what kind of distortion existing compression algorithms generate; (2) how annoying a human observer finds certain types of distortion; (3) what are the possibilities of mathematically analyzing the distortion. A set of test images was comp...

In the present work we study image quality evaluation from an empirical point of view and want to find out (1) what kind of distortion existing compression algorithms generate; (2) how annoying a human observer finds certain types of distortion; (3) what are the possibilities of mathematically analyzing the distortion. A set of test images was comp...

The distortion caused by image compression is typically determined by a
pixelwise measure such as mean square error. However, these types of
measures do not consider global artifacts, like blockiness or blurring.
In the present paper an attempt is made to model the distortion by a
blockwise measure. A three-parameter distortion function is proposed...

this paper we propose a new hierarchical BTC algorithm (HBTC-VQ). The algorithm competes well in respect to bit rate, MSE, and speed. A hierarchy of blocks is considered in the algorithm and vector quantization is applied when the distortion caused by the VQ is expected to be small. We consider different design alternatives while constructing the c...

Block truncation coding (BTC) is a lossy moment preserving quantization method for compressing digital gray-level images. Its advantages are simplicity, fault tolerance, the relatively high compression efficiency and good image quality of the decoded image. Several improvements of the basic method have been recently proposed in the literature. In t...

In the present paper we study genetic algorithms for the codebook generation problem of vector quantization. There are two different approaches to the problem: a codebook-based and a partition-based. Both these approaches can be derived from the optimality criteria of GLA. From these, the codebook- based approach is clearly superior to the partitio...

## Citations

... Recent studies have indicated that playing games against other people is more fun and exciting than playing alone[3]. Currently, most multiplayer video games are designed assuming that all players essentially share the same technology[9]. That is, their gameplay relies on interconnected PCs, mobile devices, etc. Past studies have shown the use of mobile devices as a uniquely capable adjunct to games that are currently played on computers or as a stand-alone gaming platform not related to computers[7]. ...

... The split based initial centroid method combines all data points into one cluster and splits until the K clusters are found. The splitting process used standard deviation (Franti et al., 1997), bisecting KM (Steinbach et al., 2000), tri-level KM (Yu et al., 2017), and so on. ...

... A randomized procedure is considered in [9], in which a random center and a random cluster are swapped. Other deterministic procedures are proposed [10,11,11,[29][30][31]. ...

Reference: A Geometric Approach to $k$-means

... The PNN is an attractive approach for clustering because of its conceptual simplicity and relatively good results [7]. It has also been combined with k-means clustering as proposed in Ref. [8], or used as a component in more sophisticated optimization methods. ...

... There are many variations of this algorithm . For example, Virmajoki (2002) Shi et al. (2005) and Wang et al. (2007a) propose algorithms that closely resemble the main idea of the meanshift method. Although data sharpening procedure proposed by Choi and Hall (1999) is originally designed to reduce bias by pushing data points at the boundary a bit closer to the center, the movements also resemble the one in the mean-shift method. ...

... Specifically, this codebook is generated by the implementation of bit probability as formulated by BF = [f v f h f d ] , namely Bernoulli bit pattern codebook. Let B = {B 1 , B 2 , … , B k } be the Bernoulli bit pattern codebook generated from binary vector quantization with soft centroid method (Fränti and Kaukoranta 1999) and composed of B k binary codewords. To acquire the BF of an image, the bit pattern indexing process matches DDBTC bitmap images of block bm(i, j) with the codeword B i in the Bernoulli bit pattern codebook. ...

... The implementation takes O(N 2 ) time in total. Further speed-up can be achieved by using lazy update of the merge cost values [18], and by reducing the amount of work caused by the distance calculations [7]. All the variants cited above give either asymptotic or relative improvement in the time complexity but they do not provide any improvements in the clustering quality. ...

... Several swap-based clustering algorithms have been considered in literature. Deterministic swap selects the prototype to be swapped as the one that increases the objective function value f least [5][6][7], or by merging two existing clusters [8,9] following the spirit of agglomerative clustering. The new location of the prototype can be chosen by considering all possible data vectors [7], splitting an existing cluster [7,10], or by using some heuristic such as selecting the cluster with the largest variance [5]. ...

Reference: Efficiency of random swap clustering

... We now describe how to adopt the table data structure [15] to maintain the above manipulations. Fig. 4 ...

... We identify new vectors using Bayesian predictive identi cation. By this we mean that we identify a new vector z with the class C j ; j = 1; 2; : : : ; k+1, which maximizes p(zja j ; C j ) j : (30) If expression (30) assumes its maximum for j = k + 1, a new class is formed. The identi cation as de ned above can be equivalently based on the discriminant functions l j (z) = d X i=1 w ij jz i ? a ij j + b j + log(t j + 1); j = 1; : : : ; k; (31) and l k+1 (z) = log( ? ...