Search

In video captioning, the description of a video usually relies on the domain to which the video belongs. Typically, the videos belong to wide range domains such as sports, music, news, cooking, etc. In many cases, a video can be associated with more than one domain. In this paper, we propose an approach to video captioning that uses domain-specific decoders. We build a domain classifier to obtain the estimates of probabilities of a video belonging to different domains. For each video, we identify the top − k domains based on the estimated probabilities. Each video in the training data set is shared in training the domain-specific decoders of top−k labels obtained from the domain classifier. The domain-specific decoders use the domain-specific semantic tags for generating captions. The proposed approach uses the Temporal VLAD for preprocessing the features extracted from 2D-CNN and 3D-CNN features. The preprocessed features provide better feature representation of the videos. The effectiveness of the proposed approach is demonstrated through the results of experimental studies on Microsoft Video Description (MSVD) corpus and MSR-VTT dataset.

Image Captioning is generating a human-readable textual description or a sentence about an image. The proposed semantic driven CNN-LSTM architecture comprises of the feature extraction process, semantic keywords extraction, facial recognition, and encoder-decoder LSTM networks. A pre-trained CNN is used to extract features from an image. A semantic keywords extraction module is used to identify the objects present in the image. The objects identified are labeled as the semantic tags present in the image. It increases the efficiency of captions in describing the objects and inclusion of these semantic labels in the captions. The LSTM based language model generates the captions by producing one word at a time. The facial recognition system identifies and recognizes the celebrity faces in the images, we have collected faces dataset which has facial images 232 celebrities. The instances of the person in the sentence were replaced with their names and personalized captions were generated. The Bilingual evaluation understudy (BLEU) and METEOR scores were generated to calculate the precision of generated captions.

Objectives: The vital goals of the face recognition research was for the use of optimization algorithms in an efficient biometric process. The algorithms are compared for biometrics on face recognition to implement a cost-effective and efficient continuous authentication system to substantially reduction the risk of fraud. Methods/Statistical Analysis: By trial comes to fruition, the proposed system upgraded the face recognition precision procured by the processed content was around 2.97% which is greater above Eigen face procedure when compared to standard PCA strategy. In addition, above trial comes about describing that the given strategy has increased and efficient optimization exactness. Findings: To enhance persistent authentication handle amid online procedure of different continuous application and retain both the client’s hard and soft biometric data and persistently verifying whether the individual using the terminal is as an indistinguishable substantial client from the one login toward the start. It is a financially savvy and productive persistent authentication framework to significantly lessen the danger of extortion. The resistance against ridiculing is inspected. Application/Improvements: Improvements are done on differentiating PSO and ACO and E (pso+aco)- based approaches for highlight assurance. The typical acknowledgment rate of E (pso+aco) was good when compared to PSObased component decision. On analyzing it was found that, number of parts needed in E (pso+aco) was not that much required for acknowledgment utilizing PSO

Background/Objectives: In worldwide, Glaucoma is basically a second major retinal disease that results in permanent blindness. Loss of Retinal Nerve Fiber Layer (RNFL) is the outcome of glaucoma disease. RNFL thickness is evaluated as a function of spatial information from Optical Coherence Tomography (OCT) images is a significant diagnostics indicator intended for glaucoma disease. However, due to factors such as low image contrast, speckle noise, high spatial resolution, exact compression of OCT is complex. To solve above issues, a Dual Tree Complex Wavelet Transform (DTCWT) based OCT image compression is proposed in this research work. Methods/Statistical analysis: The proposed method consists of five phases such as pre-processing, feature extraction, RNFL segmentation, glaucoma classification and OCT compression. Initially, OCT image is pre-processed for remove the speckle noise using kuan filter. Secondly, the RNFL based texture features are extracted by using Gray Level Covariance Matrix (GLCM) and the scrupulous features are chosen by Principal Component Analysis (PCA). Then, RNFL in OCT is segmented by Improved Artificial Bee Colony (IABC) clustering algorithm. After that the glaucoma is classified as normal, medium and severe by Modular Neural Network (MNN). Finally, DTCWT is used to compress the OCT image. Results: Experimental results show that the proposed MNN is efficient for detecting glaucoma compared with the existing detection algorithms.

Image captioning is gaining attention due to the recent developments in the deep neural architectures. But the gap between semantic concepts and the visual features is a major challenge in image caption generation. In this paper we have developed a method to use both visual features and semantic features for the caption generation. We discuss briefly about the various architectures used for visual feature extraction and Long Short Term Memory (LSTM) for caption generation. An object recognition model has been developed to identify the semantic tags in the images. These tags are encoded along with the visual features for the captioning task. We have developed an Encoder-Decoder architecture using the semantic details along with the language model for the caption generation. We evaluated our model with standard datasets like Flickr8k, Flickr30k and MSCOCO using standard metrics like BLEU and METEOR.

In the biomedical field, prognosis prediction of high-risk diseases which are common among the adult patient remains a challenge. The risk factor and symptoms of the high-risk diseases like diabetes, blood pressure, and hyperlipidemia are inflated in adult patients and typically lead to medical complications like fatal death. Despite many studies, the prognosis prediction still remains a challenge. In this paper, a model is proposed for prognosis prediction of high-risk diseases by designing LSTM (long short-term memory)-based RNN (recurrent neural network) deep learning network. This work learns diagnostic histories of patients and predicts the incidence of high-risk diseases by using ICD-10 codes. The experimental results depict that the LSTM-RNN provides better performance with respect to precision, recall, and F1-score when compared to RNN-GRU. Further, the proposed model also investigates the effects of the learning parameters with respect to performance.

Brain tumor is a fatal disease among the human community and affects people irrespective of their age, gender and locality. In this paper, a semi-automated approach is considered to extract the tumor region from the benchmark brain MRI known as the BRATS 2015. A two-step procedure by integrating the tri-level thresholding and level-set segmentation is implemented to extract the tumor section from the MRI. Initially, Tsallis entropy assisted thresholding is implemented to enhance the tumor section; later the segmentation procedure is implemented to extract the tumor. After extracting the infected section, a relative analysis is executed with respect to the ground truth image in order to evaluate the performance of the proposed semi-automated tool. The experimental results of this work confirm that, proposed approach offers better average result for the Jaccard, Dice, sensitivity, specificity and accuracy. In future, this procedure can be considered to examine the real-time brain MRI obtained from the clinics.

Automatic detection and classification of different types of arrhythmias by analyzing the ECG signal is essential for diagnosis of cardiac defects. Thus the focus of this paper is to propose an efficient classifier to classify the ECG signal into different arrhythmias such as Normal, VE, VF, LBBB and APB to support in diagnosis. Initially Histogram features are extracted from RR interval, PT interval, PR interval, TT interval, ST interval, QT interval of the ECG signal. Then the Real Coded Genetic Algorithm (RCGA) is applied to identify the optimal feature set from the extracted histogram features. After that the Multilayer feed forward neural network (MFFNN) evolved using Particle Bee approach. The performance of the proposed RCGA-PBNN has been evaluated in terms of classification accuracy, sensitivity and specificity using the MIT-BIH arrhythmias ECG Database.

In this research letter, a low resolution (LR) satellite image is enhanced to high resolution (HR) image using various wavelet techniques. One of the key factors is image resolution. Image RE is the method of modifying a portrait so that output portrait is more superior to the novel one for meticulous application. There are many techniques for portrait enhancement in image processing. Mainly three important wavelet techniques are used i.e., Method1 is Discrete wavelet transform (DWT), Method2 is Discrete wavelet transform — Stationary wavelet transforms (DWT-SWT), and Method3 is Dual tree complex wavelet transform (DT-CWT). A low resolution portrait is decomposed into four sub-band (SB's) i.e., Low-Low (SB 1), low-high (SB 2), high-low (SB 3), and high-high (SB 4) using DWT, DWT-SWT, and DT-CWT. The high frequency (HF) SB's are interpolated using bicubic interpolation and combined with interpolated low resolution portrait in case of DWT-SWT and DT-CWT. SB 1, SB 2, SB 3, SB 4 are interpolated using bicubic interpolation and combined with difference of input portrait and interpolated Low frequency SB portrait in case of DWT. Inverse DWT, Inverse DWT-SWT, and Inverse DT-CWT wavelet transform are applied respectively. The output is high resolution (HR) portrait. Image Resolution enhancement loses HF components. So DWT technique is used to preserve HF components. DWT generates artifacts. These artifacts are reduced by DT-CWT. All these three techniques are applied on NOAA-19-HRPT satellite images. The quantitative PSNR, RMSE, and CC are calculated for satellite portraits and compared for these three wavelet techniques.

This study adapted an improved algorithm based on Artifical Bee Colony Optimization. It is not possible to justify that all the rules generated by fuzzy based apriori algorithm produce optimum result. Thus optimization of the result generated was carried out by Fuzzy Apriori algorithm using Fuzzy Artifical Bee Colony Optimization (FABCO), it's worth noting that a significant findings were revealed. FABCO is used for optimization of rules to get the best classification accuracy. The proposed method was compared with the traditional Artifical bee colony optimization and the particle swarm optimization. The current work proved a better classification performance compared to un-pruned rules.

Segmenting and tracking multiple humans is a challenging problem in complex situations in which extended occlusion, shadow and/or reflection exists. We tackle this problem with a 3D model-based approach. This method includes two stages, segmentation (detection) and tracking. Human hypotheses are generated by shape analysis of the foreground blobs using a human shape model. The segmented human hypotheses are tracked with a Kalman filter with explicit handling of occlusion. Hypotheses are verified while they are tracked for the first second or so. The verification is done by walking recognition using an articulated human walking model. We propose a new method to recognize walking using motion template and temporal integration. Experiments show that our approach works robustly in very challenging Sequences.

In Medical Diagnosis, Magnetic Resonance Image (MRI) plays a momentous role. MRI is based on the physical and chemical principles of Nuclear Magnetic Resonance (NMR), a technique used to gain information about the nature of molecules. Retrieving a high quality MR Image for a medical diagnosis is critical. So denoising of Magnetic Resonance (MR) images and making them easy for human understanding form is a challenge. This research work presents an efficient Hybrid Abnormal Detection Algorithm (HADA) to detect the abnormalities in any part of the human body by MRIs. The proposed technique includes five stages: Noise Reduction, Smoothing, Feature Extraction, Feature Reduction and Classification. The proposed algorithm has been implemented and Classification accuracy of 98.80% has been achieved. The result shows that the proposed technique is robust and effective compared to other recent works. The system developed using the proposed algorithm will be a good computer aided diagnosis and decision making system in healthcare.

In machine learning, classification refers to a step by step procedure for designating a given piece of input data into any one of the given categories. There are many classification problem occurs and need to be solved. Different types are classification algorithms like tree-based, rule-based, etc are widely used. This work studies the effectiveness of Rule-Based classifiers for classification by taking a sample data set from UCI machine learning repository using the open source machine learning tool. A comparison of different rule- based classifiers used in Data Mining and a practical guideline for selecting the most suited algorithm for a classification is presented and some empirical criteria for describing and evaluating the classifiers are given.

Travelers, sailors and research people working in sea environment are facing lot of challenges. One such challenge is salt water crocodile. In general, crocodiles are seen in fresh water like lake and river. Nowadays it is migrated to salt water. These salt water crocodiles are serious threat to travelers and sailors. This issue is addressed through solution like Ocean Observation System (OOS), which integrates video frame system with target detection. Development of OOS is a challenging task due to various image processing modules like background occlusion, object deformation, scale changes, etc. Detection algorithms using global feature descriptor such as colour, failed to produce reliable detection, when the background image has similar colour. In case of motion pictures, reliability of detection becomes less while using local feature descriptors like Histogram of Oriented Gradient (HOG), Scale-Invariant Feature Transform (SIFT) etc., As target detection through such single feature representation does not produce reliable results, we propose a Multi-Feature Joint Descriptor (MFJD) method in this paper. An optimization of crocodile detection process is obtained through the implementation of two different MFJD and by comparing their performance on detection rate. The feature descriptors considered in the proposed MFJD algorithms includes Colour, HOG and SIFT. The simulated results ensure the reliability of detection achieved through MFJD detection process.

a day's object tracking is more difficult and tricky to surveillance in real time. This proposed work deals with the tracking of moving object in a sequence of frames and it also determines the velocity of the object. In this work algorithms are developed for improving the image quality, segmentation, feature extraction and for identifying the velocity. The algorithms developed are implemented and evaluated using MATLAB. The image quality of the video frame is obtained by applying certain noise removal filters. Next, identifying the moving objects from the portion of the video frame is performed using the background subtraction technique based on frame difference. The object tracking is performed by optical flow with Bayesian boosting algorithm method on detected object in each frame as a feature extraction method. There are several papers describing the accomplishment of optical flow. Some results are adequate, but in many projects, there are restrictions. In most preceding applications, because the camera is typically static, it is simple to apply optical flow to identify the moving targets in a scene and get their trajectories. When the camera moves a global motion will be added to the local motion, which complicates the issue. In this work we used the combination of boosting algorithm called Bayesian boosting is adopted to improve the performance of Optical flow (OFTBB). The distance traveled by the object is determined using its centroid pixel. It is calculated by using the Euclidean distance formula. Then the velocity of the object is calculated by finding the object moved in distance in a sequence of frames with respect to the video frame rate. Keywordstracking, Preprocessing, Segmentation, Feature Extraction, Tracking and Detection.

In this work we express technological strategies using mobile captured symptoms of cotton leaf spot images and classify the diseases using neural network. The system has been trained to achieve intelligent farming for rural area farmers, including early recognition of diseases in grows, selective fungicide application,etc..This research work proposes an automatic image preprocessing techniques. At first, the captured images are processed for improvement. Other edge detectors presented in earlier works can detect edges on different size objects. In this Research work, a homogeneity operator can take the difference of the center pixel and a pixel that is two or three pixels away. The major objective of this Research work is to use Homogeneity-based edge detector segmentation, which takes the result of any edge detector and divides it by the average value of the area. This work has been implemented in the real time software and produces best results.

Video mining has grown as an energetic research area and given incremental concentration in recent years due to impressive and rapid raise in the volume of digital video databases. The aim of this research work is to find out new objects in videos. This work proposes a novel approach for video mining using LIM based clustering technique and self organizing maps to recognize novelty in the frames of video sequence. The proposed work is designed and implemented on MATLAB. It is tested with the sample videos and provides promising results. And it is suitable for day to day video mining applications and object detection systems including remote video surveillance in defense for national and international border tracking.

With the rapid development of digital cameras, we have witnessed great interest and promise in automatic image annotation as a hot research field. Automatic image annotation is an effective method to resolve the problem of “Semantic Gap”. Automatic Image annotation is a challenging problem when a label is provided for the entire training image only instead of the object region. To eliminate labeling ambiguity, image categorization and object localization should be performed simultaneously. Discriminative Multiple Instance multi-label Learning (MIML) can be used for this task by regarding each image as a bag and sub-windows in the image as instances. Learning a discriminative Multi-instance classifier requires an iterative solution. In each round, positive sub-windows for the next round should be selected. With standard approaches, selecting only one positive sub-window per positive bag may limit the search space for global optimum; meanwhile, selecting all temporal positive sub-windows may add noise into learning. We select a subset of sub-windows per positive bag to avoid those limitations. Our Proposed EMID algorithm is able to take the correlation among instances, correlation among labels, and correlation between instances and labels simultaneously, and provides a very rich representation and learning potential. Experimental results demonstrate that our approach outperforms previous discriminative MIML approaches and standard categorization approaches.

Image fusion is the method of combining relevant information from two or more images into a single image resulting in an image that is more informative than the initial inputs. Methods for fusion include discrete wavelet transform, Laplacian pyramid based transform, curvelet based transform etc. These methods demonstrate the best performance in spatial and spectral quality of the fused image compared to other spatial methods of fusion. In particular, wavelet transform has good time-frequency characteristics. However, this characteristic cannot be extended easily to two or more dimensions with separable wavelet experiencing limited directivity when spanning a one-dimensional wavelet. This paper introduces the second generation curvelet transform and uses it to fuse images together. This method is compared against the others previously described to show that useful information can be extracted from source and fused images resulting in the production of fused images which offer clear, detailed information.

In Medical Diagnostic, Magnetic Resonance Images play a major role. Magnetic Resonance images are normally corrupted by random noise from the measurement process complicating the automatic feature extraction and analysis of clinical data. Because of this reason noise removal methods have been customarily applied to improve MR image quality. This work proposed a new scheme based on applying a series of filters, each used to modify the estimate into greater agreement, so that the output converges to a stable estimate providing noise free image. In this work, we have introduced a novel hybrid filter to reduce random noise in MR images by the combination of Kernel, Sobel and Low-pass (KSL) filtering techniques. The proposed method has been implemented using Matlab and compared with related state of art methods over synthetic and real clinical MR images showing a superior performance in all cases analyzed.

database. It is commonly used in a marketing, surveillance, fraud detection and scientific discovery. In data mining, machine learning is mainly focused as research which is automatically learnt to recognize complex patterns and make intelligent decisions based on data. Nowadays traffic accidents are the major causes of death and injuries in this world. Roadway patterns are useful in the development of traffic safety control policy. This paper deals with the some of classification models to predict the severity of injury that occurred during traffic accidents. I have compared Naive Bayes Bayesian classifier, AdaBoostM1 Meta classifier, PART Rule classifier, J48 Decision Tree classifier and Random Forest Tree classifier for classifying the type of injury severity of various traffic accidents. The final result shows that the Random Forest outperforms than other four algorithms.

Finding motif in biosequences is the most important primitive operation in computational biology. There are many computational requirements for a motif discovery algorithm such as computer memory space requirement and computational complexity. To overcome the complexity of motif discovery, we propose an alternative solution integrating genetic algorithm and Fuzzy Art machine learning approaches for eliminating multiple sequence alignment process. Problem statement: More than a hundred methods had been proposed for motif discovery in recent years, representing a large variation with respect to both algorithmic approaches as well as the underlying models of regulatory regions. The aim of this study was to develop an alternative solution for motif discovery, which benefits from both data mining and genetic algorithm, and which at the same time eliminates the cost caused by use of multiple sequence alignment. Approach: Genetic algorithm based probabilistic Motif discovery model was designed to solve the problem. The proposed algorithm was implemented using Matlab and also tested with large DNA sequence data sets and synthetic data sets. Results: Results obtained by the proposed model to find the motif in terms of speed and length are compared with the existing method. Our proposed method finds Length of 11 in 18 sec and length of 15 in 24 sec but the existing methods finds length of 11 in 34 sec. Compare to other techniques the proposed one was outperforms the popular existing method. Conclusion: In this study, we proposed a model to discover motif in large set of unaligned sequences in considerably minimum time. Length of motif was also long. The proposed algorithm will be implemented using Matlab and was tested with large DNA sequence data sets and synthetic data sets.

Face Recognition system finds many applications in surveillance and human computer interaction systems. As the applications using face recognition systems are of much importance and demand more accuracy, more robustness in the face recognition system is expected with less computation time. In this paper, a Hybrid approach for face recognition combining Gabor Wavelet and Linear Discriminant Analysis (HGWLDA) is proposed. The normalized input grayscale image is approximated and reduced in dimension to lower the processing overhead for Gabor filters. This image is convolved with bank of Gabor filters with varying scales and orientations. LDA, a subspace analysis techniques are used to reduce the intra-class space and maximize the inter-class space. The techniques used are 2-dimensional Linear Discriminant Analysis (2D-LDA), 2-dimensional bidirectional LDA ((2D)2LDA), Weighted 2-dimensional bidirectional Linear Discriminant Analysis (Wt (2D)2 LDA). LDA reduces the feature dimension by extracting the features with greater variance. k-NearestNeighbour (k-NN) classifier is used to classify and recognize the test image by comparing its feature with each of the training set features. The HGWLDA approach is robust against illumination conditions as the Gabor features are illumination invariant. This approach also aims at a better recognition rate using less number of features for varying expressions. The performance of the proposed HGWLDA approaches is evaluated using AT&T database, MIT-India face database and faces94 database. It is found that the proposed HGWLDA approach provides better results than the existing Gabor approach.

In this era, an emerging field in the data mining is data stream mining. The current research technique of the data stream is classification which mainly focuses on concept drift data. Mining drift data with the single classifier is not sufficient for classifying the data. Because of the high dimensionality and failure to get processed within considerable time and memory, false alarm rate is high, and classification accuracy result is low. In chapter, a proposed genetic-based intuitionistic fuzzy version of k-means has been introduced for grouping interdependent features. The proposed method achieves improvement in classification accuracy and perhaps in selecting the least number of features which show the way to simplification of learning task. The experiment shows that the advocated method performs well when compared with existing methods.

In this paper, an approach for face recognition combining multi-resolution analysis and transform domain analysis is proposed. Face Recognition system find its use in many applications such as authentication, surveillance, human-computer interaction systems etc. As the applications using face recognition systems are of much importance and demand more accuracy, more robustness in the face recognition system is expected with less computation time. In the proposed ComGW-DCT approach, features are extracted using a combination of Gabor filters and Discrete Cosine Transform (DCT). The normalised input grayscale image is approximated and reduced in dimension to lower the processing overhead for Gabor filters. This image is convolved with bank of Gabor filters with varying scales and orientations. Further DCT technique is adapted to reduce the feature space dimension. DCT extracts low frequency components of the Gabor wavelet thus resulting in the compression of Gabor features. For classification, k-Nearest Neighbour (k-NN) classifier is used to recognise the test image by comparing with each of the training set features. The ComGW-DCT approach is robust against illumination conditions as the Gabor features are illumination invariant. This algorithm also aims at better recognition rate using less number of features for varying expressions without affecting the computation time. The results of the proposed system are evaluated using AT&T database and MIT-India face database.

Fuzzy Association rule mining is an essential topic in Information retrieval mining field. It produces all important Fuzzy association rules between attributes in the dataset because large data set records considered as transactions. Each transaction consists of set of attributes. Fuzzy based Apriori algorithm produces all important Fuzzy association rules between attributes in the dataset. On the basis of the Fuzzy association rule mining and Fuzzy Apriori algorithm, the proposed work adapted an improved algorithm based on Artificial Bee Colony Optimization. It is not possible to justify that all the rules generated by Fuzzy based Apriori algorithm produce optimum result. Thus optimization of the result generated was carried out by Fuzzy Apriori algorithm using Fuzzy Artificial bee colony optimization (FABCO), its worthy noting that a significant findings were revealed. FABCO is used for optimization of rules to get the best classification accuracy. The proposed method was compared with the traditional artificial bee colony optimization and the particle swarm optimization. The current work proved a better classification performance compared to un-pruned rules.

Abandoned object discovery using tracking based approaches are often become unpredictable in complex surveillance videos due to noise, occlusions, lighting changes, and other factors. This proposed paper presents a new structure to strongly and proficiently detect abandoned objects based on background subtraction. In this proposed system, the background is modelled by Gaussian mixture. In order to hold complex situations, numerous enhancements are implemented for noise removal, rapid lighting transform adaptation, fragment reduction, and maintaining a stable update rate for video streams with different frame rates. In order to substantiate this proposed approach the object detection method using Intuitionistic logic based on block matching techniques has been used. The experimental results obtained were tested on benchmark video sequences. The obtained results are very promising in terms of robustness and effectiveness.

The neural network is a very useful tool for approximation of a function, time series prediction, classification, and pattern recognition. If there is found to be a non-linear relationship between input data and output data, it is difficult to analyse the system. A neural network is very effective to solve this problem. This chapter studies the applied neural network model in relation to clearance sales outshopping behaviour. Since neural network theory can be applied effectively to this case, the authors have used neural network theory to recognise the retail area satisfaction and loyalty. To measure the impact among the retail area attributes, retail area satisfaction, and retail area loyalty, the authors have used the neural network model. In this chapter, they have treated twenty seven factors as the input signals into the input layer. Therefore, they find the weights between nodes in the relationship between the value of all twenty seven factors and the retail area satisfaction and loyalty. The development of the model by retail area attributes, and their interpretation, was facilitated by a collection of data across three trading areas. This neural network modeling approach to understand clearance sales outshopping behaviour provides retail managers with information to support retail strategy development.

Video transmission plays a very important role in traffic applications. Noise can be a big offence in affect-ing encoding efficiency because it can be present throughout an entire application. Noise has the technical definition for various anomalies and unnecessary variations that get built-in into a video signal. Noise re-duction enables better video quality at lower bit rates by making the source look better and decrease the video complication prior to the any process. In this proposed method we adapted the spatial video denois-ing methods, where image noise are reduced and are is applied to each frame individually. Since there is a great deal of removing noise from video content, this paper has been devoted to noise detection and filter-ing methods that aims the removing unwanted noise without affecting the clarity of scenes which contains necessary information and rapid movement. The aim of this work is to produce the exact intensity information of segmentation’s neighborhood relationships [1]. In this paper, foreground based segmentation; fuzzy c-means clustering segmentation is compared with the proposed method fuzzy c – means segmentation based on color. This was applied in the video frame to segment various objects in the current frame. The proposed technique is a commanding method for image segmentation and it works for both single and multiple featured data for spatial information. Strong techniques were introduced for finding the number of components in an image. The results done experimentally shows that the proposed segmentation approach generates good quality segmented frames. This paper deals with efficient analysis of noise removal techniques and enhancing the segmentation in video frames.

Intruders attack both commercial and corporate distributed systems successfully. The problem of intruders has become vital. The most effective resistance today is the use of Intrusion Detection Systems. An intrusion detection system analysis all aspects of network activities in order to identify the existence of unusual patterns that may represent a network or system attack made by intruders attempting to compromise a system. This paper brings an idea of applying data mining algorithms to the intrusion detection system. Performance of various tree based classifiers like Decision Stump, BF Tree, ID3, J48, LAD, Random Tree, REP Tree, Random Forest and Simple Cart algorithms are compared and the experimental study shows that the Random Forest algorithm outperforms than other algorithms in terms of accuracy, specificity and sensitivity and Time.

A feature is defined as a role of one or more measurements, the values of some irrefutable assets of an object, computed so that it quantifies some significant characteristics of an object. Here in this paper we can have presented a clearer view of how feature extraction takes place on the basis of color, texture, and shape in video processing. The variation that occurs during feature extraction is also shown. The Iteration in which the objects are identified from a specified location is also considered. The proposed algorithm shows that the objects are identified correctly in a lower number of iterations itself when compared to the traditional algorithms.

It is very much essential for the multimedia information organization to provide accurate and scalable solutions to map low-level perceptual features to high-level semantics. Therefore automatic and efficient annotation of images is needed for rapid content based retrieval and indexing; it alleviates the disadvantage of any manual annotation. The proposed system for pattern matching and annotation from large image databases has been given based on the combination of Fractal Transform and gentle AdaBoost algorithm. This technique involves two main stages in classification phase wherein first, we make use of gentle AdaBoost algorithm as it is best suited for object detection task and also has lower computational complexity. Next, a mathematical representation is associated to the images of the database, this representation is a set of function parameters resulting from a dedicated fractal interpolation scheme, and used as an index by a retrieval algorithm. Proposed algorithm works completely in the Fractal transform parameter space of both images and patterns, to obtain performances well-matched with an interactive search. In this paper, we also try to overcome the orientation, scaling and class imbalance problem in image annotation by choosing an over sampling method for learning the classifier. Experimental results of IFSMOTE shows higher prediction quality, and performs better than the classical SVM, SMOTE and FSMOTE.

Video transmission plays a very important role in traffic applications. Noise can be a big offence in affecting encoding efficiency because it can be present throughout an entire application. Noise has the technical definition for various anomalies and unnecessary variations that get built-in into a video signal. Noise reduction enables better video quality at lower bit rates by making the source look better and decrease the video complication prior to the any process. In this proposed method we adapted the spatial video denoising methods, where image noise are reduced and are is applied to each frame individually. Since there is a great deal of removing noise from video content, this paper has been devoted to noise detection and filtering methods that aims the removing unwanted noise without affecting the clarity of scenes which contains necessary information and rapid movement. The aim of this work is to produce precise segmentation of images using intensity information along with neighborhood relationships. Most of the results of color image segmentation are based on gray level image segmentation methods with different color representations were published. Image segmentation techniques such as histogram threshold, clustering in segmentation, region growing, edge detection, fuzzy methods, and neural networks can be extended to color images.

Data mining has been popularly recognized as an important way to mine useful information from large volumes of data that are noisy, fuzzy & random. Intrusion detection has become an efficient tool against network attack because they allow network administrator to detect vulnerability. Existing IDS techniques includes high false positive and false negative rate. Data mining using IDS reduces the number of false alarm rate. So, here some of the clustering algorithms like k means, hierarchical and Fuzzy C Means have been implemented to analyze the detection rate over KDD CUP 99 dataset. Based on evaluation result, FCM outperforms in terms of both accuracy and computational time.

Network security has become an important issue due to the evolution of internet. It brings people not only together but also provides huge potential threats. Intrusion detection technique is considered as the immense method to deploy networks security behind firewalls. An intrusion is defined as a violation of security policy of the system. Intrusion detection systems are developed to detect those violations. Due to the effective data analysis method, data mining is introduced into IDS. This paper brings an idea of applying data mining algorithms to intrusion detection database. Performance of various rule and function based classifiers like Part, Ridor, NNge, DTNB, JRip, Conjunctive Rule, One R, Zero R, Decision Table, RBF, Multi Layer Perception and SMO algorithms are compared and result shows that SMO classification algorithm performs well in terms of accuracy, specificity and sensitivity. The performance of the model is measured using 10- fold cross validation.

Proposed Research work exposes, a advance computing technology has been developed to help the farmer to take superior decision about many aspects of crop developed process. Suitable evaluation and diagnosis of crop disease in the field is very critical for the increased production. Foliar is the major important fungal disease of cotton and occurs in all growing Indian cotton regions. In this work we express Technological Strategies using mobile captured symptoms of Cotton Leaf Spot images and categorize the diseases using neural network. The classifier is being trained to achieve intelligent farming, including early detection of disease in the groves, selective fungicide application, etc. This proposed work is based on Image Edge detection Segmentation techniques in which, the captured images are processed for enrichment first. Then R, G, B color Feature image segmentation is carried out to get target regions (disease spots). Later, image features such as boundary, shape, color and texture are extracted for the disease spots to recognize diseases and control the pest recommendation.

Recently the progress in technology and flourishing applications open up new forecast and defy for the image and video processing community. Compared to still images, video sequences afford more information about how objects and scenarios change over time. Quality of video is very significant before applying it to any kind of processing techniques. This paper deals with two major problems in video processing they are noise reduction and object segmentation on video frames. The segmentation of objects is performed using foreground segmentation based and fuzzy c-means clustering segmentation is compared with the proposed method Improvised fuzzy c – means segmentation based on color. This was applied in the video frame to segment various objects in the current frame. The proposed technique is a powerful method for image segmentation and it works for both single and multiple feature data with spatial information. The experimental result was conducted using various noises and filtering methods to show which is best suited among others and the proposed segmentation approach generates good quality segmented frames.

This paper presents a novel image annotation framework for domains with large numbers of images. Automatic image annotation is such a domain, by which a computer system automatically assigns metadata in the form of captioning or keywords to a digital image. This application of computer vision method is used in image retrieval system to organize and locate images of interest from a database. Many techniques have been proposed for image annotation in the last decade that has given realistic performance on standard datasets. In this work, we introduce an innovative hybrid model for image annotation that treats annotation as a retrieval problem. The proposed technique utilizes low level image features and a simple combination of basic distances using JEC to find the nearest neighbors of a given image; the keywords are then assigned using SVM approach which aims to explore the combination of three different methods. First, the initial annotation of the data using flat wise and axis wise methods, and that takes the hierarchy into consideration by classifying consecutively its instances through position wise method. Finally, we make use of pair wise majority voting between methods by simply summing strings in order to make a final annotation. The result of the proposed technique shows that this technique outperforms the current state of art methods on the normal datasets.

Automatic image annotation is the process by which a computer system automatically assigns metadata in the form of captioning or keywords to a digital image. This application of computer vision technique is used in image retrieval system to organize and locate images of interest from a database. Many techniques have been proposed for image annotation in the last decade that gives reasonable performance on standard datasets. In this work, we introduce an innovative hybrid model for image annotation that treats annotation as a retrieval problem. The proposed technique utilizes low level image features and a simple combination of basic distances using JEC to find the nearest neighbors of a given image; the keywords are then assigned using SVM approach which aims to explore the combination of three different methods. First, the initial annotation of the data using two known methods, and that takes the hierarchy into consideration by classifying consecutively its instances; finally, we make use of pair wise majority voting between methods by simply summing strings in order to produce a final annotation. The proposed technique results show that this method outperforms the current state of art methods on the standard datasets.

In Medical Diagnosis, Magnetic Resonance Images play a significant role. This work presents a hybrid technique to detect the abnormal parts in the magnetic resonance images of a human. The proposed technique consists of five stages namely, Noise reduction, Smoothing, Feature extraction, Dimensionality reduction or Feature reduction, and Classification. First stage obtains the noise reduced MR image using KSL (Kernel-Sobel-Low pass) Filter. In the second stage, smoothing is done by histogram equalization. Third stage obtains the features related with MRI images using discrete wavelet transformation. In the fourth stage, the features of magnetic resonance images have been reduced using principles component analysis to the more essential features. At last stage, the classifier based on K-means has been used to classify subjects as normal or abnormal MRI human images. Classification accuracy of 98.80% has been obtained by the proposed algorithm. The result shows that the proposed technique is robust and effective compared with other recent works.

To extract implicit knowledge and data relationships from the audio and audio similarity measure, this paper uses the audio mining techniques. A model for audio clustering and classification technique is proposed. Neural networks are used for classifying the data. The working prototype of the Music classification system has been developed and tested in MATLAB 6.5 using the signal Processing Toolbox under Windows XP operating system in a normal Pentium range of desktop computer.

A simple technique employed for the restoration of the obliterated painted registration number on a motor vehicle involved in a case of murder is described.

This Proposed Work exposes, a advance computing technology that has been developed to help the farmer to take superior decision about many aspects of crop development process. Suitable evaluation and diagnosis of crop disease in the field is very critical for the increased production. Foliar is the major important fungal disease of cotton and occurs in all growing Indian regions. In this work we express new technological strategies using mobile captured symptoms of cotton leaf spot images and categorize the diseases using HPCCDD Proposed Algorithm. The classifier is being trained to achieve intelligent farming, including early Identification of diseases in the groves, selective fungicide application, etc. This proposed work is based on Image RGB feature ranging techniques used to identify the diseases (using Ranging values) in which, the captured images are processed for enhancement first. Then color image segmentation is carried out to get target regions (disease spots). Next Homogenize techniques like Sobel and Canny filter are used to Identify the edges, these extracted edge features are used in classification to identify the disease spots. Finally, pest recommendation is given to the farmers to ensure their crop and reduce the yeildloss.

Recent investigates on performance appraisal of video segmentation methods is principally focused on the progress of objective facts of advantage. The majority of the computer visualization and image analysis problems necessitate a segmentation stage in order to discover objects or divide the image into regions. This contribution describes using color based fuzzy c-means method in image segmentation. We have compared the performance of the proposed segmentation approach with two other segmentation methods foreground subtraction method and fuzzy c means segmentation on grey images. The proposed work produces better segmentation result than other two.