Abstract and 1 Introduction 2 Related Work 3 Image Compression Module 4 System Architecture 5 Experimental Results 6 Conclusions A Compression Results References 2 Related Work Popular remote display protocols offer high-resolution sessions, multimedia stream remoting, multi-monitor support, dynamic object compression, USB redirection, drive mapping and more. Microsoft Remote Desktop Protocol (RDP), VMwares PC-over-IP (PCoIP) and Citrixs HDX are the most commonly used, but there are other protocols from companies such as Ericom and Hewlett-Packard. Teradici sells hardware solutions based on PCoIP technology [10] and the Thin Client Internet Computing protocol is also applicable for remote displays [11]. But these protocols still have drawbacks; great performance requires a lot of bandwidth and energy, which can clog the network [35]. Plus, if the system is running with low CPU usage, protocols will hog bandwidth and weaken end user performance. So, an efficient system needs to cope with unstable network conditions, to operate in real time, and to satisfy the users’ demands. Almost all of these protocols have the disadvantages that they do not consider the network environment and are inconvenient to use with real-time applications such as video playback. Previous work has included the development of an adaptive encoding application to solve display performance problems encountered when using RDP [12]. The authors of another work have described problems that arise in the case of video conferencing and have developed a real-time camera control system to address the problems [13]. To reduce the size of images transferred over the network to improve video streaming quality, compression technology has been one of the most important contributions. Some encoding and decoding formats such as motion MPEG [14], [16] and H.264 AVC [15], [17] have been developed in hybrid remote display protocol. The challenges for compressing images have been summarized previously [18]. Various compression techniques have been developed for image compression [37], [38]. In this work, we focus on the k-means clustering for encoding the image data. This method is probably the most popular technique of representativebased clustering. Cluster analysis [44] is a major data analysis method that is widely used for many practical applications in emerging areas. Clustering is the process of partitioning a given set of objects into disjoint clusters. This is done in such a way that objects in the same cluster are similar whereas objects belonging to different clusters differ considerably with respect to their attributes [48]. The k-means algorithm [45], [46] is effective in producing clusters for many practical applications. But the original k-means algorithm is very computationally complex, especially for large data sets. Moreover, this algorithm yields different clusters depending on the random choice of initial centroids. For example, previous work has included the description of a 2D image compression scheme using k-means in a real-time multiprocessor system [39]. Another work presented a novel k-means algorithm for compressing images, including exploration of the strategy of applying statistical parameters for choosing the initial seeds [40]. Despite the efficiency of the latter scheme, it still has drawbacks like a priori fixation of the number of clusters and random selection of initial seeds. Various researchers have made several attempts to improve the performance of k-means clustering [41], [42], [43]. In addition, hierarchical clustering algorithms also usually apply in image processing. But as the number of records increase the performance of hierarchical algorithm goes decreasing and time for execution increased. K-mean algorithm also increases its time of execution but as compared to hierarchical algorithm its performance is better. As a general conclusion, k-means algorithm is good for large dataset and hierarchical is good for small datasets [6]. Therefore k-means is most suitable for image processing in the video streaming scheme. Beside, developing a model that efficiently provides QoS control is presently a challenging problem. Many researches have addressed the problem of providing QoS guarantees to various network applications and clients. Most of them have focused on servers without considering network delays [19], [20], on individual network routers [29], [30], [31], or on clients with assumptions of QoS supports in networks [32]. Some recent work focused on end-to-end QoS guarantees in network cores [33], [34], with the aim of guaranteeing QoS as measured from server-side network edges to client-side network edges. To solve this problem, previous approaches need to be revised to allow their application to a cloud environment in a manner that responds to various user behaviors and network conditions. In the present work, k-means clustering is used to reducing the size of one frame from the video. We then develop a scheme for the analysis of historical data as a means to consider network status and make a decision regarding the optimal compression ratio for QoS control. Eventually, we evaluate our system model using actual measurement results resulting from the processing of images from a realistic image dataset. This paper is available on arxiv under CC BY 4.0 DEED license. (1) Huu-Quoc Nguyen, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and quoc@khu.ac.kr;
(2) Tien-Dung Nguyen, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and ntiendung@khu.ac.kr;
(3) Van-Nam Pham, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and nampv@khu.ac.kr;
(4) Xuan-Qui Pham, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and pxuanqui@khu.ac.kr;
(5) Quang-Thai Ngo, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and nqthai@khu.ac.kr;
(6) Eui-Nam Huh, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and johnhuh@khu.ac.kr. Abstract and 1 Introduction Abstract and 1 Introduction 2 Related Work 2 Related Work 3 Image Compression Module 3 Image Compression Module 4 System Architecture 4 System Architecture 5 Experimental Results 5 Experimental Results 6 Conclusions 6 Conclusions A Compression Results A Compression Results References References 2 Related Work Popular remote display protocols offer high-resolution sessions, multimedia stream remoting, multi-monitor support, dynamic object compression, USB redirection, drive mapping and more. Microsoft Remote Desktop Protocol (RDP), VMwares PC-over-IP (PCoIP) and Citrixs HDX are the most commonly used, but there are other protocols from companies such as Ericom and Hewlett-Packard. Teradici sells hardware solutions based on PCoIP technology [10] and the Thin Client Internet Computing protocol is also applicable for remote displays [11]. But these protocols still have drawbacks; great performance requires a lot of bandwidth and energy, which can clog the network [35]. Plus, if the system is running with low CPU usage, protocols will hog bandwidth and weaken end user performance. So, an efficient system needs to cope with unstable network conditions, to operate in real time, and to satisfy the users’ demands. Almost all of these protocols have the disadvantages that they do not consider the network environment and are inconvenient to use with real-time applications such as video playback. Previous work has included the development of an adaptive encoding application to solve display performance problems encountered when using RDP [12]. The authors of another work have described problems that arise in the case of video conferencing and have developed a real-time camera control system to address the problems [13]. To reduce the size of images transferred over the network to improve video streaming quality, compression technology has been one of the most important contributions. Some encoding and decoding formats such as motion MPEG [14], [16] and H.264 AVC [15], [17] have been developed in hybrid remote display protocol. The challenges for compressing images have been summarized previously [18]. Various compression techniques have been developed for image compression [37], [38]. In this work, we focus on the k-means clustering for encoding the image data. This method is probably the most popular technique of representativebased clustering. Cluster analysis [44] is a major data analysis method that is widely used for many practical applications in emerging areas. Clustering is the process of partitioning a given set of objects into disjoint clusters. This is done in such a way that objects in the same cluster are similar whereas objects belonging to different clusters differ considerably with respect to their attributes [48]. The k-means algorithm [45], [46] is effective in producing clusters for many practical applications. But the original k-means algorithm is very computationally complex, especially for large data sets. Moreover, this algorithm yields different clusters depending on the random choice of initial centroids. For example, previous work has included the description of a 2D image compression scheme using k-means in a real-time multiprocessor system [39]. Another work presented a novel k-means algorithm for compressing images, including exploration of the strategy of applying statistical parameters for choosing the initial seeds [40]. Despite the efficiency of the latter scheme, it still has drawbacks like a priori fixation of the number of clusters and random selection of initial seeds. Various researchers have made several attempts to improve the performance of k-means clustering [41], [42], [43]. In addition, hierarchical clustering algorithms also usually apply in image processing. But as the number of records increase the performance of hierarchical algorithm goes decreasing and time for execution increased. K-mean algorithm also increases its time of execution but as compared to hierarchical algorithm its performance is better. As a general conclusion, k-means algorithm is good for large dataset and hierarchical is good for small datasets [6]. Therefore k-means is most suitable for image processing in the video streaming scheme. Beside, developing a model that efficiently provides QoS control is presently a challenging problem. Many researches have addressed the problem of providing QoS guarantees to various network applications and clients. Most of them have focused on servers without considering network delays [19], [20], on individual network routers [29], [30], [31], or on clients with assumptions of QoS supports in networks [32]. Some recent work focused on end-to-end QoS guarantees in network cores [33], [34], with the aim of guaranteeing QoS as measured from server-side network edges to client-side network edges. To solve this problem, previous approaches need to be revised to allow their application to a cloud environment in a manner that responds to various user behaviors and network conditions. In the present work, k-means clustering is used to reducing the size of one frame from the video. We then develop a scheme for the analysis of historical data as a means to consider network status and make a decision regarding the optimal compression ratio for QoS control. Eventually, we evaluate our system model using actual measurement results resulting from the processing of images from a realistic image dataset. This paper is available on arxiv under CC BY 4.0 DEED license. This paper is available on arxiv under CC BY 4.0 DEED license. available on arxiv (1) Huu-Quoc Nguyen, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and quoc@khu.ac.kr; (2) Tien-Dung Nguyen, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and ntiendung@khu.ac.kr; (3) Van-Nam Pham, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and nampv@khu.ac.kr; (4) Xuan-Qui Pham, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and pxuanqui@khu.ac.kr; (5) Quang-Thai Ngo, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and nqthai@khu.ac.kr; (6) Eui-Nam Huh, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and johnhuh@khu.ac.kr. (1) Huu-Quoc Nguyen, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and quoc@khu.ac.kr; (2) Tien-Dung Nguyen, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and ntiendung@khu.ac.kr; (3) Van-Nam Pham, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and nampv@khu.ac.kr; (4) Xuan-Qui Pham, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and pxuanqui@khu.ac.kr; (5) Quang-Thai Ngo, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and nqthai@khu.ac.kr; (6) Eui-Nam Huh, Department of Computer Engineering, Kyung Hee University, 1 Seocheon, Giheung, Yongin, Gyeonggi, South Korea and johnhuh@khu.ac.kr.

Part of HackerNoon's growing list of open-source research papers, promoting free access to academic material.

Abstract

Cluster

Microsoft

System

Focusing on the K-Means Clustering for Encoding the Image Data

About Author

Comments

TOPICS

THIS ARTICLE WAS FEATURED IN

Related Stories

A New Architecture With Flexible QoS Control for Virtual Desktop Infrastructure

Introduction to Clustering in Machine Learning: Types, Algorithms, and Applications

Learn K-Means Clustering by Quantizing Color Images in Python

Mastering K-Means: Data Clustering Simplified

Optimizing Novel Class Discovery with NCD Spectral Clustering and k-means

How to Fit an Elephant in a Spreadsheet

A New Architecture With Flexible QoS Control for Virtual Desktop Infrastructure

Introduction to Clustering in Machine Learning: Types, Algorithms, and Applications

Learn K-Means Clustering by Quantizing Color Images in Python

Mastering K-Means: Data Clustering Simplified

Optimizing Novel Class Discovery with NCD Spectral Clustering and k-means

How to Fit an Elephant in a Spreadsheet

Light-Mode

Classic

Newspaper

Minty

Dark-Mode

Neon Noir

Minty

HN StartUps