Papers by Dr Thyagharajan K K
Objective: In recent times, the demand for high data rates is ever increasing in any wireless net... more Objective: In recent times, the demand for high data rates is ever increasing in any wireless network environment. Long Term Evolution-Advanced (LTE-A) is the latest 4G technology which is developed based on 3GPP specifications. Our main objective in this proposed research wok is to analyze the various packet scheduling algorithms for downlink real time data and present their scheduling metrics. Methods: In this review we considered more recent scheduling algorithms which are QoS aware and with specific focus on real time traffic classes such as VoIP, Video Streaming, Interactive Gaming and mobile video conference. Conclusion: Our most significant observation from this review is that any packet scheduling algorithm for downlink real time data should be QoS aware so that it is readily deployable in the present day multimedia networks. The scheduling schemes should also consider the latest technologies such as Carrier Aggregation (CA) and Multi Input and Multi Output (MIMO) Applications: The presented review will help the researchers and academicians to develop more efficient scheduling schemes for real time applications for smart phone users with better quality of experience and efficient radio resource management.
Objectives: To classify the satellite images into different land use/land cover classes such as w... more Objectives: To classify the satellite images into different land use/land cover classes such as water, building, cropland, forest, etc, to monitor the environmental impacts. Method: In this paper, images are grouped into various clusters using a novel SVD trace function clustering algorithm. The clustered samples are used as a training set in a novel unsupervised Ensemble Minimization Learning algorithm (EML) for classification. The main aim of using EML is to classify the forest, vegetative land patterns, build up area in rural and urban areas with the use of best accuracy rate. Finding: Our proposed methods provides 90.56% classification rate with low error rate. This EML applies multinomial probit model and ensembles simulated data set and improves the learning of nonlinear relationships between the classified attributes. Multinomial probit model is used to bring all the related possible segmented values to fall into one single category, thus increasing the classification accuracy. Our proposed methods experimented with three different real data sets. The experimental results indicate that our proposed unsupervised model outperforms than the previous techniques. Application: It could be using for land use/land cover change detection, under water object identification, coastal area monitoring, etc. Improvement: In future it could be apply in video data and could be improve the classification accuracy also.
In this paper some of the different techniques used to localize the lips from the face are discus... more In this paper some of the different techniques used to localize the lips from the face are discussed and compared along with its processing steps. Lip localization is the basic step needed to read the lips for extracting visual information from the video input. The techniques could be applied on asymmetric lips and also on the mouth with visible teeth, tongue & mouth with moustache. In the process of Lip reading the following steps are generally used. They are, initially locating lips in the first frame of the video input, then tracking the lips in the following frames using the resulting pixel points of initial step and atlast converting the tracked lip model to its corressponding matched letter to give the visual information. A new proposal is also initiated from the discussed techniques. The lip reading is useful in Automatic Speech Recognition when the audio is absent or present low with or without noise in the communication systems. Human-Computer communication also will require speech recognition.
Colors in an image provides tremendous amount of information. Using this color information images... more Colors in an image provides tremendous amount of information. Using this color information images can be segmented, analyzed, labeled and indexed. In content based image retrieval system, color is one of the basic primitive features used. In Prevalent Color Extraction and indexing, the most extensive color on an image is identified and it is used for indexing. For implementation, Asteroideae flower family image dataset is used. It consist of more than 16,000 species, among them nearly 100 species are considered and indexed by dominating colors. To extract the most appealable color from the user defined images, the overall color of an image has to be quantized. Spatially, quantizing the color of an image to extract the prevalent color is the major objective of this paper. A combination of K-Mean and Expectation Minimization clustering algorithm called hidden-value learned K-mean clustering quantization algorithm is used to avoid the over clustering behavior of K-Mean algorithm. The experimental result shows the marginal differences between these algorithms.
The fused image derived from multimodality, multi-focus, multi-view and multidimensional, for rea... more The fused image derived from multimodality, multi-focus, multi-view and multidimensional, for real world applications in the field of medical imaging, remote sensing, satellite imaging, machine vision etc., are gaining much attention in the recent research. Therefore, it is important to validate the fused image in different perspectives such as information, edge, structure, noise and contrast for quality analysis. From this aspect, the information of fused image should be better than the source images without loss of information and false information. This survey is focused on analyzing the various quantitative metrics that are used in the literature to measure the enhanced information of fused image when it is compared to the source/reference images. The objective of this study is to group or classify the metrics under different categories such as information, noise, error, correlation and structural similarity measures for discussion and analysis. In reality, the calculated metric values are useful in determining the suitable fusion technique of the particular dataset with its required perspective as an outcome of the fusion process.
We proposed a method for Human action Recognition. It is based on the construction of a set of te... more We proposed a method for Human action Recognition. It is based on the construction of a set of templates for each activity. Each template is constructed based on the Accumulated Motion Image of the Video. Each template contains where motion has occurred in the video. FFT Transform is applied to each template. A 3D Spatiotemporal Volume is generated for each class. A Single action Maximum average Correlation height Filter is generated for each class. The filter is applied to the test video and using the threshold the actions are classified. The experiments are conducted on Weizmann dataset.
Image fusion is the process of combining multiple images into a single image without distortion o... more Image fusion is the process of combining multiple images into a single image without distortion or loss of information. The techniques related to image fusion are broadly classified as spatial and transform domain methods. In which, the transform domain based wavelet fusion techniques are widely used in different domains like medical, space and military for the fusion of multimodality or multi-focus images. In this paper, an overview of different wavelet transform based methods and its applications for image fusion are discussed and analysed.
H.264 Advanced Video Coding (AVC) was prolonged to Scalable Video Coding (SVC). SVC executes in d... more H.264 Advanced Video Coding (AVC) was prolonged to Scalable Video Coding (SVC). SVC executes in different electronics gadgets such as personal computer, HDTV, SDTV, IPTV, and full-HDTV in which user demands various scaling of the same content. The various scaling is resolution, frame rate, quality, heterogeneous networks, bandwidth, and so forth. Scaling consumes more encoding time and computational complexity during mode selection. In this paper, to reduce encoding time and computational complexity, a fast mode decision algorithm based on likelihood mode decision (LMD) is proposed. LMD is evaluated in both temporal and spatial scaling. From the results, we conclude that LMD performs well, when compared to the previous fast mode decision algorithms. The comparison parameters are time, PSNR, and bit rate. LMD achieve time saving of 66.65% with 0.05% detriment in PSNR and 0.17% increment in bit rate compared with the full search method.
The modality based medical image analysis for brain tumor studies is gaining much attention in re... more The modality based medical image analysis for brain tumor studies is gaining much attention in recent era due to the reason that a large amount of images has to be processed with clinical data to analyze different types of outcomes. The outcome may be diagnosis of diseases, computer assisted surgery, to predict the growth of a disease after a period, retrieving similar cases, drug discovery etc,. In the field of Computer Aided Disease Diagnosis (CAD), classification technique plays a major role in identifying the image/tissue as normal or abnormal. This paper focuses on the statistical and texture based Brain image classification using Support Vector Machine for different modalities (MRI, PET/SPECT) and also fused images of MRI and PET/SPECT. The result shows that the fused images are classified correctly (94%) with minimum false positive error rate than the single modality images.
H.264 / AVC expansion is H.264 / SVC which is applicable in environments that demand video stream... more H.264 / AVC expansion is H.264 / SVC which is applicable in environments that demand video streaming. This paper delivers an algorithm to shorten computational complexity and extend coding efficiency by determining the mode speedily. In this writing, the authors talk a fast mode resolution algorithm with less complexity unlikely the traditional joint scalable video model (JSVM). Their algorithm end mode hunt by a probability model defined. This model is address for both intra-mode and inter-mode predictions of base layer and enhancement layers in a macro block (MB). The estimated rate distortion cost (RDC) for modes among layers is custom to determine the best mode of each MB. The experimental results show that the authors' algorithm realizes 26.9% of encoding time when compared with the JSVM reference software with smallest reduction in peak signal to noise ratio (PSNR).
Information is said to be valid and knowledgeable if all the subjects and objects with respect to... more Information is said to be valid and knowledgeable if all the subjects and objects with respect to a particular domain are all arranged in a quantitative manner. To design a semantic information retrieval system, all these subjects and objects with respect to its predicate have to be defined in the given database. This can be done by Ontology Engineering. The main scope of this paper is to build an intellectual semantic search framework, which has the combination of both textual semantic and visual semantics. As the number of available information increases, the use of classical ontology for information retrieval process will not provide satisfactory results. So, in this paper we have explained the construction of fuzzy ontology to identify the action of football game. By giving image as query, the system with the help of fuzzy ontology rule setup will identify the action performed in the image.
Information retrieval is one of the most common web service used. Information is knowledge. In ea... more Information retrieval is one of the most common web service used. Information is knowledge. In earlier days one has to find a resource person or resource library to acquire knowledge. But today just by typing a keyword on a search engine all kind of resources are available to us. Due to this mere advancement there are trillions of information available on net. So, in this era we are in need of search engine which also search with us by understanding the semantics of given query by the user. One such design is only possible only if we provide semantic to our ordinary HTML web page. In this paper we have explained the concept of converting an HTML page to RDFS/OWL page. This technique is incorporated along with natural language technology as we have to provide the Hyponym and Meronym of the given HTML pages. Through this automatic conversion the concept of intelligent information retrieval is framed.
Multimedia is one of the important means of communication channel for mankind. Due to the advance... more Multimedia is one of the important means of communication channel for mankind. Due to the advancement in technology and enormous growth in civilizations, there is a tremendous amount of multimedia data available today. This obviously calls for the development of some technique for retrieving these data. This paper will give an overview of one such ontology-based image retrieval system for the Asteroideae Flower Family domain. Initially, the image ontology, with respect to the selected low-level features such as prevalent color descriptor, edge histogram descriptor, color layout descriptor and Texton was created. Thus created ontology can be used as a back-end system for semantic-based image retrieval system.
Scalable Video Coding is an extension of H.264/AVC, which has better coding efficiency and scalab... more Scalable Video Coding is an extension of H.264/AVC, which has better coding efficiency and scalability in terms of temporal, spatial and quality levels and the Scalability is done based on the value of Quantization Parameter (QP). This paper aims to compare the quantization parameter value for Base Layer and Enhancement Layer. The bit rates and the PSNR generated under each layer of a YUV sequence for the assigned QP value in Base Layer and Enhancement Layer is compared using JSVM 9.19.5 SVC Reference Software Model, and the experimental results shows that the Base Layer assigned with small QP value than the Enhancement Layer generates better PSNR and reduced bit rate.
Video Transmission over wireless network suffers a lot of discrepancies in Quality of Experience ... more Video Transmission over wireless network suffers a lot of discrepancies in Quality of Experience at the end user. Wireless LAN is adopted widely everywhere even at home because of its simpler installation, cheaper, flexible and reliability. A novel approach Dynamic Transmission Opportunity over Weight Based Scheduling (DTWBS) is proposed to transmit video at end user with enhanced QoS. Dynamic bandwidth allocation is performed over measuring the future traffic demands based on the piggybacked requests made by the QoS enabled stations (QSTA) to QoS enabled Access Point (QAP) for the variable bit rate video. Prior to this, H.264 standard video is being coded with a specified fixed Group of Pictures (GOP) having no two keyframes are aligned to occur within 2 seconds. The offered bandwidth from the QAP is categorized to a dynamic percentage of weights based on different class of frames/slices with their significant information. The Scheduler in our proposed scheme polls the more weight video frame/slice which has more significant information followed by a frame/slice with less significant information. Hence, this approach dynamically uses the bandwidth with improved Quality of Experience at the end user without the loss of less significant frames/slices which achieves fairness.
H.264/AVC extension is H.264/SVC which is applicable for environment that demands video streaming... more H.264/AVC extension is H.264/SVC which is applicable for environment that demands video streaming. This paper presents an algorithm to reduce computation complexity and maintain coding efficiency by determining the mode quickly. Our algorithm terminates mode search by a probability model for both intra-mode and inter-mode of lower level and higher level layers in a Macro Block (MB). The estimated of Rate Distortion Cost (RDC) for modes among layers is used to determine best mode of each MB. This algorithm achieves about 26.9% of the encoding time when compared with JSVM reference software with minimal degradation in PSNR.
The mere existence of natural living thing can be studied and analyzed efficiently only by Ontolo... more The mere existence of natural living thing can be studied and analyzed efficiently only by Ontology, where each and every existence are concern as entities and they are grouped hierarchically via their relationship. This paper deals the way of how an image can be represented by its feature Ontology though which it would be easier to analyze and study the image automatically by a machine, so that a machine can visualize an image as human. Here we used the selected MPEG 7 visual feature descriptor and Texton parameter as entity for representing different categories of images. Once the image Ontology for different categories of images is provided image retrieval would be an efficient process as through ontology the semantic of image is been defined.
Information is knowledge. In earlier days one has to find a resource person or resource library t... more Information is knowledge. In earlier days one has to find a resource person or resource library to acquire knowledge. But today just by typing a keyword on a search engine all kind ofresources are available to us. Due to this mere advancement there are trillions of information available on net. So, in this era we are in need of search engine which also search with us by understanding the semantics of given query by the user. One such design is only possible only if we provide semantic to our ordinary HTML web page. In this paper we have explained the concept of converting an HTML page to RDFS/OWL page. This technique is incorporated along with natural language technology as we have to provide the Hyponym and Meronym of the given HTML pages.
In the existing research of mammogram image classification either clinical data or image features... more In the existing research of mammogram image classification either clinical data or image features of specific type is considered along with the supervised classifiers such as Neural Network (NN) and Support Vector Machine (SVM). This paper considers automated classification of breast tissue type as benign or malignant using Weighted Feature Support Vector Machine (WFSVM) through constructing the precomputed kemel t'unction by assigning more weight to relevant features using the principle of maximizing deviations. Initially, MIAS dataset of mammogram images is divided into training and test set, then the preprocessing techniques such as noise removal and background removal are applied to the input images and the Region of Interest (ROI) is identified. The statistical fbatures and texture features are extracted from the ROI and the clinical features are obtained directly from the dataset. The extracted features of the training dataset are used to construct the weighted features and precomputed linear kernel for training the WFSVM, from which train model file is created. Using this model file the kernel matrix of test samples are classified as benign or malignant. These analysis shows that the texhrre features are resulted with better accuracy than the other features with WFSVM and SVM. However the number of support vectors created in WFSVM is less than the SVM classifier.
Past researchers have shown maximum recognition rates given the static images and techniques to r... more Past researchers have shown maximum recognition rates given the static images and techniques to recognize face and related features of the face. Yet, these research though contribute and motivate greatly to building effective future systems, fail to address the temporal dynamics of the face to enhance the system's training. In this paper, analyzing the facial expression on a given face from an image automatically and produces the result stating the emotion on the subject's face. Face is recognized using the skin and chrominance of the extracted image and the image is cropped. Expressions on the face are determined using the localization of points called Action Units (AUs) internally without labeling them. Though AUs are found to be effective, most expressions on the face have shown to overlap these points thereby curbing the recognition. Using a mapping technique, the extracted eyes and mouth are mapped together. Illumination on an image plays a vital role in highlighting the portrait and therefore is a barrier when extracting the facial features. This is a delimiter while analyzing the face. This limitation is removed and automatically corrected using a Color Constancy Algorithm with minkowski norms. The experimental results show better face detection rate under variable luminance levels. The system was tested against a collection of faces both containing single face images and multiple faces in a scene. We achieved a recognition rate of 60% when detecting in a multiple face image.
Uploads
Papers by Dr Thyagharajan K K