radoslav vargic - Academia.edu (original) (raw)

Papers by radoslav vargic

2017 4th International Conference on Control, Decision and Information Technologies (CoDIT), 2017

In this paper we analyse a basic hierarchical mask creation methods for image coding using salien... more In this paper we analyse a basic hierarchical mask creation methods for image coding using saliency maps. For saliency maps (SM) based image coding we use specific extension of SPIHT algorithm called SM SPIHT that extends region of interest encoding to encoding with individual weight of importance for each pixel in image using the form of saliency map. This approach is proved to be effective. In this article, we compare basic hierarchical mask creation methods and provide one new method that outperforms all previous methods.

In this paper we analyze the relationship between integer Lifting scheme and Rounding transform a... more In this paper we analyze the relationship between integer Lifting scheme and Rounding transform as means to compute the wavelet transform in signal processing area. We bring some new results which better describe relationship, reversibility and equivalence of integer lifting scheme and rounding transform concept.

We propose a method for coding of segmented images based on shape adaptive wavelet transform appr... more We propose a method for coding of segmented images based on shape adaptive wavelet transform approach. Wavelet transform is computed directly on particular image segments without extension outside the boundaries. We adapted the transform method from [1], where it was used for compression of image segments but less effective zero tree coding was involved. Though used approach does not yet outperform the non-segmented methods in the sense of MSE, the visual quality of image segments is comparable. We propose also an alternative adaptation with comparative results, which can be used with classical DWT implementations.

2018 25th International Conference on Systems, Signals and Image Processing (IWSSIP)

This research aims to increase quality of experience for video consumers by adaptive delivery of ... more This research aims to increase quality of experience for video consumers by adaptive delivery of multimedia content. We summarize ways of detecting salient regions in observed scene and current state of dynamic adaptive streaming over HTTP (DASH). We propose to extend DASH video delivery system, based on saliency information gathered using eye tracking. This extension enables to enhance the quality in certain regions in video and is particularly suitable for adaptive multimedia content delivery in 5G Networks.

2018 25th International Conference on Systems, Signals and Image Processing (IWSSIP), 2018

In this paper a role of virtual reality in self-directed learning is introduced and discussed. Wi... more In this paper a role of virtual reality in self-directed learning is introduced and discussed. Within a H2020 project Newton a new platform for distance learning based on student centric model is designed and realized. Fields like learning management system, virtual laboratories, augmented reality, assessment method and multimodal system are important parts of the designed platform and are included in this paper. Main focus is oriented to game-based learning using virtual reality (VR) in education. As an example, game-based VR application describing the functionality of firewall is discussed. “Firewall” application is going to be used in one of the pilot projects for testing the philosophy and Newton platform in education. We describe the proposed integration of the VR applications to the learning platform and propose the usage assessment.

2018 International Symposium ELMAR, 2018

This research aims to increase quality of experience for video consumers by introducing an approa... more This research aims to increase quality of experience for video consumers by introducing an approach with saliency-based video foveation coupled with compression algorithm. This approach uses eye tracking information to build the saliency map and allows multiple usage scenarios. The approach can be used in single or multiple viewer environments. We evaluate the approach and provide results based on subjective measurements. The results confirm that the proposed approach for video compression is competitive and can deliver better quality of experience than standard video compression algorithms.

2017 4th International Conference on Control, Decision and Information Technologies (CoDIT), 2017

Most of the information processed by human brain comes from visual sources. That is why visual in... more Most of the information processed by human brain comes from visual sources. That is why visual information is very important in human communication and decision making process. It is even more important for people whose other senses are negatively impaired. There are many people all over the world who rely on sign language to communicate with others. One of these languages is the American Sign Language (ASL). Many systems and algorithms that attempt to translate this communication into text with various success rates have been developed, but these systems cannot analyze each frame of video sequence in its full size. That is why an algorithm capable of tracking the gesturing hand and identifies key frames containing entire signs is highly desirable and important. Even more so considering how much are video channels, especially wireless video channels, susceptible to disturbances and noise.

2011 18th International Conference on Systems, Signals and Image Processing, 2011

This article presents the task of speaker identification in a closed group. It discusses main ste... more This article presents the task of speaker identification in a closed group. It discusses main steps of the identification process ranging from the proper speech features to the classification methods and statistical signal processing. However, its main focus is on tuning the final system using KNN classification method by setting up the number of neighbors, and reducing the feature vector dimension by PCA and LDA not only to speed up but possibly improve the overall performance. By selecting eligible number of neighbors a 6% improvement in the recognition was reached. Moreover, application of both PCA and LDA reduced the feature vector dimension by more than 50% while slightly increasing the recognition accuracy.

We present a new method to estimate the Hurst parameter. The method exploits the form of the auto... more We present a new method to estimate the Hurst parameter. The method exploits the form of the autocorrelation function for second-order self-similar processes and is based on one-pass digital filtration. We compare the performance and properties of the new method with that of the most common methods.

In this paper we provide an unequal error protection enhancement for SM SPIHT based image coding ... more In this paper we provide an unequal error protection enhancement for SM SPIHT based image coding and transmission. SM SPIHT coding uses SPIHT algorithm as basis and employs the saliency maps (SM) to better capture the significance of each image pixel based on the assumed importance for the viewer. The saliency maps are proven extension of the SPIHT. In this paper, we extend this concept for the unequal error protection (UEP) suitable for progressive image transmission, compare the results and show advantages of the proposed approach.

This paper deals with evaluation of self-similarity in the data gathered from Electronic Tolling ... more This paper deals with evaluation of self-similarity in the data gathered from Electronic Tolling System using Hurst parameter. Selected basic and derived data related to position of vehicles in tolled networks gathered by tolling systems are evaluated. The basic characteristics are exploited and results for service triggering processes are stated. As it was shown the system traffic data do not show self-similarity traces, however localized traffic features provide higher presence of self-similarity that should be taken in account in the designing process.

The paper provides a concept and architecture of the Virtual SDN and NFV Laboratory and its integ... more The paper provides a concept and architecture of the Virtual SDN and NFV Laboratory and its integration with NEWTELP platform - a learning platform developed with the EU Horizon 2020 NEWTON Project. The Virtual SDN and NFV Laboratory was proposed as a virtual laboratory for teaching and research activities in the field of Software Defined Networking (SDN) and Network Function Virtualization (NFV) technologies. The paper presents the concept, implementation and testing of integration of Virtual SDN and NFV laboratory with NEWTELP platform.

The article extends Vector Quantization (VQ) binary error generation model by a soft boundary con... more The article extends Vector Quantization (VQ) binary error generation model by a soft boundary concept utilizing Gaussian Mixture Model (GMM). Although VQ based binary error model provides superior modeling for common stochastic characteristics of digital channels it depends on unknown distance, suffers from lack of data, hard boundary space division, etc. To alleviate these drawbacks GMM based approach was applied to binary error generation process. The model was evaluated by several statistical measures against real data acquired in a wireless sensor network. The experiments using GMM show lower averaged and minimal statistical distances observed across different modeling settings.

Abstract. Image oversegmentation creates small, compact, and irregularly shaped regions subject t... more Abstract. Image oversegmentation creates small, compact, and irregularly shaped regions subject to further clustering. Consideration of texture characteristics can improve the resulting quality of the clustering process. Existing methods based on an orthogonal transform into frequency domain can extract texture features of arbitrarily shaped regions only from inscribed rectangles. We propose a method for extracting texture features of entire arbitrarily shaped image regions using orthogonal transforms. Furthermore, we introduce a mathematically correct method for unifying spectral dimensions that is necessary for accurate comparison and classification of spectra with different dimensions. The proposed method is particularly suitable for classifying areas with periodic and quasiperiodic textures. Our approach exploits the texture periodification property of certain orthogonal transforms that is based on insertion of zeros into the spectrum. We identified some of those orthogonal tran...

This paper presents a new method for compression of audio and speech signals based on sinusoidal ... more This paper presents a new method for compression of audio and speech signals based on sinusoidal modeling with added wavelet based coding of the residual signal. Wavelets are introduced as effective tool for representation and compression of atonal and transient signals. The method is proposed for usage in speech databases. The presented method is evaluated by the means of PSNR and PESQ/ODG. The results are compared with performance of common methods. The results show that the presented method provides a promising tool for speech compression and speech databases.

In this paper, we present current state and aims of our ongoing research aimed at advanced intera... more In this paper, we present current state and aims of our ongoing research aimed at advanced interactive multimedia and mulsemedia delivery in 5G networks. We summarize the underlying and necessary properties of 5G networks, architecture of SDN/NFV and their usage in the multimedia and mulsemedia delivery. Advanced topics as content adaptation and user identification are discussed. Preliminary results regarding algorithms for content adaptation are presented. We show the advance of usage of AR/VR headsets as effective terminal for immersive application for this kind of delivery and discuss the advantages and disadvantages. We propose corresponding system architecture for advanced interactive multimedia delivery in 5G networks.

2018 25th International Conference on Systems, Signals and Image Processing (IWSSIP), 2018

This research aims to increase quality of experience for video consumers by adaptive delivery of ... more This research aims to increase quality of experience for video consumers by adaptive delivery of multimedia content. We summarize ways of detecting salient regions in observed scene and current state of dynamic adaptive streaming over HTTP (DASH). We propose to extend DASH video delivery system, based on saliency information gathered using eye tracking. This extension enables to enhance the quality in certain regions in video and is particularly suitable for adaptive multimedia content delivery in 5G Networks.

2017 IEEE 11th International Conference on Application of Information and Communication Technologies (AICT)

Modern medical diagnostic systems have greatly contributed to the increase in survival rate of pa... more Modern medical diagnostic systems have greatly contributed to the increase in survival rate of patients suffering from illnesses and to lengthening of average lifespan. Some diagnostic devices, such as Magnetic Resonance, can detect illnesses that may not exhibit any symptoms yet and medical diagnostic equipment can furthermore significantly aid in confirmation of suspected, otherwise undetectable diagnoses. One of the medical fields lacking automated diagnostic tools is psychopathology, where a mental disorder presence is typically established from observations of classified symptoms and structured systematic interviews. The goal of our paper was to design and test a reliable and automated diagnostic method for detection of Schizophrenia spectrum disorders. The proposed method utilizes eye-tracker and Rorschach Inkblot Test to create a visual attention saliency map of observed subjects. These maps are then processed and analyzed using Digital Image Processing and statistical methods to determine, whether the subject exhibits signs of schizophrenia or not. The proposed approach is based on a trained classifier which separates incoming data into either healthy or schizophrenic patients. Of course, these results are only indicative and the final diagnosis rests in the hands of qualified specialists. High correlation of the proposed system, s diagnosis and real clinical diagnosis however proves applicability of the proposed concepts as reliable supplementary tool for Schizophrenia detection.

International Journal of Advances in Telecommunications, Electrotechnics, Signals and Systems

In this paper we analyze basic mask creation methods for intelligent image coding using saliency ... more In this paper we analyze basic mask creation methods for intelligent image coding using saliency maps. For saliency maps based image coding we use specific extension of SPIHT algorithm called SM SPIHT related to region of interest encoding but extending this approach further, ending with individual weight of importance for each pixel in image using the form of saliency map. This approach is proved to be effective. In this article we analyze impact of different basic hierarchical mask creation methods, which have impact on error separation between salient and not salient parts of the image. The results indicate that proposed mask creation method outperforms JPEG2000 based mask tree creation method.

Journal of Electronic Imaging, 2016

Abstract. Visual information is very important in human perceiving of the surrounding world. Duri... more Abstract. Visual information is very important in human perceiving of the surrounding world. During the observation of the considered scene, some image parts are more salient than others. This fact is conventionally addressed using the regions of interest approach. We are presenting an approach that captures the saliency information per pixel basis using one continuous saliency map for a whole image and which is directly used in the lossy image compression algorithm. Although for the encoding/decoding part of the algorithm, the notion region is not necessary anymore; the resulting method can, due to its nature, efficiently emulate large amounts of regions of interest with various significance. We provide reference implementation of this approach based on the set partitioning in hierarchical trees (SPIHT) algorithm and show that the proposed method is effective and has potential to achieve significantly better results in comparison to the original SPIHT algorithm. The approach is not limited to SPIHT algorithm and can be coupled with, e.g., JPEG 2000 as well.

2017 4th International Conference on Control, Decision and Information Technologies (CoDIT), 2017

In this paper we analyse a basic hierarchical mask creation methods for image coding using salien... more In this paper we analyse a basic hierarchical mask creation methods for image coding using saliency maps. For saliency maps (SM) based image coding we use specific extension of SPIHT algorithm called SM SPIHT that extends region of interest encoding to encoding with individual weight of importance for each pixel in image using the form of saliency map. This approach is proved to be effective. In this article, we compare basic hierarchical mask creation methods and provide one new method that outperforms all previous methods.

In this paper we analyze the relationship between integer Lifting scheme and Rounding transform a... more In this paper we analyze the relationship between integer Lifting scheme and Rounding transform as means to compute the wavelet transform in signal processing area. We bring some new results which better describe relationship, reversibility and equivalence of integer lifting scheme and rounding transform concept.

We propose a method for coding of segmented images based on shape adaptive wavelet transform appr... more We propose a method for coding of segmented images based on shape adaptive wavelet transform approach. Wavelet transform is computed directly on particular image segments without extension outside the boundaries. We adapted the transform method from [1], where it was used for compression of image segments but less effective zero tree coding was involved. Though used approach does not yet outperform the non-segmented methods in the sense of MSE, the visual quality of image segments is comparable. We propose also an alternative adaptation with comparative results, which can be used with classical DWT implementations.

2018 25th International Conference on Systems, Signals and Image Processing (IWSSIP)

This research aims to increase quality of experience for video consumers by adaptive delivery of ... more This research aims to increase quality of experience for video consumers by adaptive delivery of multimedia content. We summarize ways of detecting salient regions in observed scene and current state of dynamic adaptive streaming over HTTP (DASH). We propose to extend DASH video delivery system, based on saliency information gathered using eye tracking. This extension enables to enhance the quality in certain regions in video and is particularly suitable for adaptive multimedia content delivery in 5G Networks.

2018 25th International Conference on Systems, Signals and Image Processing (IWSSIP), 2018

In this paper a role of virtual reality in self-directed learning is introduced and discussed. Wi... more In this paper a role of virtual reality in self-directed learning is introduced and discussed. Within a H2020 project Newton a new platform for distance learning based on student centric model is designed and realized. Fields like learning management system, virtual laboratories, augmented reality, assessment method and multimodal system are important parts of the designed platform and are included in this paper. Main focus is oriented to game-based learning using virtual reality (VR) in education. As an example, game-based VR application describing the functionality of firewall is discussed. “Firewall” application is going to be used in one of the pilot projects for testing the philosophy and Newton platform in education. We describe the proposed integration of the VR applications to the learning platform and propose the usage assessment.

2018 International Symposium ELMAR, 2018

This research aims to increase quality of experience for video consumers by introducing an approa... more This research aims to increase quality of experience for video consumers by introducing an approach with saliency-based video foveation coupled with compression algorithm. This approach uses eye tracking information to build the saliency map and allows multiple usage scenarios. The approach can be used in single or multiple viewer environments. We evaluate the approach and provide results based on subjective measurements. The results confirm that the proposed approach for video compression is competitive and can deliver better quality of experience than standard video compression algorithms.

2017 4th International Conference on Control, Decision and Information Technologies (CoDIT), 2017

Most of the information processed by human brain comes from visual sources. That is why visual in... more Most of the information processed by human brain comes from visual sources. That is why visual information is very important in human communication and decision making process. It is even more important for people whose other senses are negatively impaired. There are many people all over the world who rely on sign language to communicate with others. One of these languages is the American Sign Language (ASL). Many systems and algorithms that attempt to translate this communication into text with various success rates have been developed, but these systems cannot analyze each frame of video sequence in its full size. That is why an algorithm capable of tracking the gesturing hand and identifies key frames containing entire signs is highly desirable and important. Even more so considering how much are video channels, especially wireless video channels, susceptible to disturbances and noise.

2011 18th International Conference on Systems, Signals and Image Processing, 2011

This article presents the task of speaker identification in a closed group. It discusses main ste... more This article presents the task of speaker identification in a closed group. It discusses main steps of the identification process ranging from the proper speech features to the classification methods and statistical signal processing. However, its main focus is on tuning the final system using KNN classification method by setting up the number of neighbors, and reducing the feature vector dimension by PCA and LDA not only to speed up but possibly improve the overall performance. By selecting eligible number of neighbors a 6% improvement in the recognition was reached. Moreover, application of both PCA and LDA reduced the feature vector dimension by more than 50% while slightly increasing the recognition accuracy.

We present a new method to estimate the Hurst parameter. The method exploits the form of the auto... more We present a new method to estimate the Hurst parameter. The method exploits the form of the autocorrelation function for second-order self-similar processes and is based on one-pass digital filtration. We compare the performance and properties of the new method with that of the most common methods.

In this paper we provide an unequal error protection enhancement for SM SPIHT based image coding ... more In this paper we provide an unequal error protection enhancement for SM SPIHT based image coding and transmission. SM SPIHT coding uses SPIHT algorithm as basis and employs the saliency maps (SM) to better capture the significance of each image pixel based on the assumed importance for the viewer. The saliency maps are proven extension of the SPIHT. In this paper, we extend this concept for the unequal error protection (UEP) suitable for progressive image transmission, compare the results and show advantages of the proposed approach.

This paper deals with evaluation of self-similarity in the data gathered from Electronic Tolling ... more This paper deals with evaluation of self-similarity in the data gathered from Electronic Tolling System using Hurst parameter. Selected basic and derived data related to position of vehicles in tolled networks gathered by tolling systems are evaluated. The basic characteristics are exploited and results for service triggering processes are stated. As it was shown the system traffic data do not show self-similarity traces, however localized traffic features provide higher presence of self-similarity that should be taken in account in the designing process.

The paper provides a concept and architecture of the Virtual SDN and NFV Laboratory and its integ... more The paper provides a concept and architecture of the Virtual SDN and NFV Laboratory and its integration with NEWTELP platform - a learning platform developed with the EU Horizon 2020 NEWTON Project. The Virtual SDN and NFV Laboratory was proposed as a virtual laboratory for teaching and research activities in the field of Software Defined Networking (SDN) and Network Function Virtualization (NFV) technologies. The paper presents the concept, implementation and testing of integration of Virtual SDN and NFV laboratory with NEWTELP platform.

The article extends Vector Quantization (VQ) binary error generation model by a soft boundary con... more The article extends Vector Quantization (VQ) binary error generation model by a soft boundary concept utilizing Gaussian Mixture Model (GMM). Although VQ based binary error model provides superior modeling for common stochastic characteristics of digital channels it depends on unknown distance, suffers from lack of data, hard boundary space division, etc. To alleviate these drawbacks GMM based approach was applied to binary error generation process. The model was evaluated by several statistical measures against real data acquired in a wireless sensor network. The experiments using GMM show lower averaged and minimal statistical distances observed across different modeling settings.

Abstract. Image oversegmentation creates small, compact, and irregularly shaped regions subject t... more Abstract. Image oversegmentation creates small, compact, and irregularly shaped regions subject to further clustering. Consideration of texture characteristics can improve the resulting quality of the clustering process. Existing methods based on an orthogonal transform into frequency domain can extract texture features of arbitrarily shaped regions only from inscribed rectangles. We propose a method for extracting texture features of entire arbitrarily shaped image regions using orthogonal transforms. Furthermore, we introduce a mathematically correct method for unifying spectral dimensions that is necessary for accurate comparison and classification of spectra with different dimensions. The proposed method is particularly suitable for classifying areas with periodic and quasiperiodic textures. Our approach exploits the texture periodification property of certain orthogonal transforms that is based on insertion of zeros into the spectrum. We identified some of those orthogonal tran...

This paper presents a new method for compression of audio and speech signals based on sinusoidal ... more This paper presents a new method for compression of audio and speech signals based on sinusoidal modeling with added wavelet based coding of the residual signal. Wavelets are introduced as effective tool for representation and compression of atonal and transient signals. The method is proposed for usage in speech databases. The presented method is evaluated by the means of PSNR and PESQ/ODG. The results are compared with performance of common methods. The results show that the presented method provides a promising tool for speech compression and speech databases.

In this paper, we present current state and aims of our ongoing research aimed at advanced intera... more In this paper, we present current state and aims of our ongoing research aimed at advanced interactive multimedia and mulsemedia delivery in 5G networks. We summarize the underlying and necessary properties of 5G networks, architecture of SDN/NFV and their usage in the multimedia and mulsemedia delivery. Advanced topics as content adaptation and user identification are discussed. Preliminary results regarding algorithms for content adaptation are presented. We show the advance of usage of AR/VR headsets as effective terminal for immersive application for this kind of delivery and discuss the advantages and disadvantages. We propose corresponding system architecture for advanced interactive multimedia delivery in 5G networks.

2018 25th International Conference on Systems, Signals and Image Processing (IWSSIP), 2018

This research aims to increase quality of experience for video consumers by adaptive delivery of ... more This research aims to increase quality of experience for video consumers by adaptive delivery of multimedia content. We summarize ways of detecting salient regions in observed scene and current state of dynamic adaptive streaming over HTTP (DASH). We propose to extend DASH video delivery system, based on saliency information gathered using eye tracking. This extension enables to enhance the quality in certain regions in video and is particularly suitable for adaptive multimedia content delivery in 5G Networks.

2017 IEEE 11th International Conference on Application of Information and Communication Technologies (AICT)

Modern medical diagnostic systems have greatly contributed to the increase in survival rate of pa... more Modern medical diagnostic systems have greatly contributed to the increase in survival rate of patients suffering from illnesses and to lengthening of average lifespan. Some diagnostic devices, such as Magnetic Resonance, can detect illnesses that may not exhibit any symptoms yet and medical diagnostic equipment can furthermore significantly aid in confirmation of suspected, otherwise undetectable diagnoses. One of the medical fields lacking automated diagnostic tools is psychopathology, where a mental disorder presence is typically established from observations of classified symptoms and structured systematic interviews. The goal of our paper was to design and test a reliable and automated diagnostic method for detection of Schizophrenia spectrum disorders. The proposed method utilizes eye-tracker and Rorschach Inkblot Test to create a visual attention saliency map of observed subjects. These maps are then processed and analyzed using Digital Image Processing and statistical methods to determine, whether the subject exhibits signs of schizophrenia or not. The proposed approach is based on a trained classifier which separates incoming data into either healthy or schizophrenic patients. Of course, these results are only indicative and the final diagnosis rests in the hands of qualified specialists. High correlation of the proposed system, s diagnosis and real clinical diagnosis however proves applicability of the proposed concepts as reliable supplementary tool for Schizophrenia detection.

International Journal of Advances in Telecommunications, Electrotechnics, Signals and Systems

In this paper we analyze basic mask creation methods for intelligent image coding using saliency ... more In this paper we analyze basic mask creation methods for intelligent image coding using saliency maps. For saliency maps based image coding we use specific extension of SPIHT algorithm called SM SPIHT related to region of interest encoding but extending this approach further, ending with individual weight of importance for each pixel in image using the form of saliency map. This approach is proved to be effective. In this article we analyze impact of different basic hierarchical mask creation methods, which have impact on error separation between salient and not salient parts of the image. The results indicate that proposed mask creation method outperforms JPEG2000 based mask tree creation method.

Journal of Electronic Imaging, 2016

Abstract. Visual information is very important in human perceiving of the surrounding world. Duri... more Abstract. Visual information is very important in human perceiving of the surrounding world. During the observation of the considered scene, some image parts are more salient than others. This fact is conventionally addressed using the regions of interest approach. We are presenting an approach that captures the saliency information per pixel basis using one continuous saliency map for a whole image and which is directly used in the lossy image compression algorithm. Although for the encoding/decoding part of the algorithm, the notion region is not necessary anymore; the resulting method can, due to its nature, efficiently emulate large amounts of regions of interest with various significance. We provide reference implementation of this approach based on the set partitioning in hierarchical trees (SPIHT) algorithm and show that the proposed method is effective and has potential to achieve significantly better results in comparison to the original SPIHT algorithm. The approach is not limited to SPIHT algorithm and can be coupled with, e.g., JPEG 2000 as well.