Charul Bhatnagar

Recognition of Indian Sign Language using feature fusion

Sign Language is the most natural and expressive way for the hearing impaired. This paper presents a method for automatic recognition of two handed signs of Indian Sign Language (ISL). The method consists of three phases: Segmentation, Feature Extraction and Recognition. The segmentation is done through Otsus algorithm. In the feature extraction phase, shape descriptors, HOG descriptors (Histogram of Oriented Gradient) and SIFT (Scale Invariant Feature Transform) feature have been fused to compute a feature vector. In the recognition phase, a multi-class Support Vector Machine (MSVM) is used for training and classifying signs of ISL. The experimental results provide evidence of the effectiveness of the proposed approach with 93% recognition rate.

LSB based non blind predictive edge adaptive image steganography

Image steganography is the art of hiding secret message in grayscale or color images. Easy detection of secret message for any state-of-art image steganography can break the stego system. To prevent the breakdown of the stego system data is embedded in the selected area of an image which reduces the probability of detection. Most of the existing adaptive image steganography techniques achieve low embedding capacity. In this paper a high capacity Predictive Edge Adaptive image steganography technique is proposed where selective area of cover image is predicted using Modified Median Edge Detector (MMED) predictor to embed the binary payload (data). The cover image used to embed the payload is a grayscale image. Experimental results show that the proposed scheme achieves better embedding capacity with minimum level of distortion and higher level of security. The proposed scheme is compared with the existing image steganography schemes. Results show that the proposed scheme achieves better embedding rate with lower level of distortion.

A framework for abandoned object detection from video surveillance

In this paper, we propose a method to detect abandoned object from surveillance video. In first step, foreground objects are extracted using background subtraction in which background modeling is done through running average method. In second step, static objects are detected by using contour features of foreground objects of consecutive frames. In third step, detected static objects are classified into human and non-human objects by using edge based object recognition method which is capable to generate the score for full or partial visible object. Nonhuman static object is analyzed to detect abandoned object. Experimental results show that proposed system is efficient and effective for real-time video surveillance, which is tested on IEEE Performance Evaluation of Tracking and Surveillance data set (PETS 2006, PETS 2007) and our own dataset.

A framework for key frame extraction from surveillance video

Key frame extraction from video is area of interest in many applications, like video analysis, video summary, semantic video indexing, video organization, and video compression. In this paper, we propose a framework for key frame extraction. The proposed framework consists of two steps: First, the size of input video shot is reduced by eliminating those frames of the shot which are not distinguishable by a human eye. Then the motion energy between the remaining frames of the input video shot is calculated and those frames are extracted as key frames in which the optical flow becomes maximum. The experimental results provide evidence of the effectiveness of the proposed approach.

A survey on automated microaneurysm detection in Diabetic Retinopathy retinal images

Automated retinal image analysis is becoming an imperative screening tool for early revealing of certain risks and diseases like Diabetic Retinopathy. Diabetic Retinopathy (DR) is the prominent cause of blindness in the world. Early detection of diabetic retinopathy can provide operative treatment. Early treatment can be conducted from detection of microaneurysms. Microaneurysms are the earliest clinical sign of diabetic retinopathy and they appear as small red spots on retinal fundus images. Microaneurysms are reddish in color with a diameter less than 125 μm. The existing trained eye care specialists are not able to screen the growing number of diabetic patients. So there is a need to develop a technique that is capable to detect microaneurysms as a part of diagnosis system, so that medical professionals are able to diagnose the stage of the disease with ease. Automated microaneurysm detection can decrease the workload of ophthalmologists and cost in DR screening system. Early automated microaneurysms detection can help in reducing the incidence of blindness. In this paper, we review and analyze the techniques, algorithms and methodologies used for the detection of microaneurysms from diabetic retinopathy retinal fundus images.

Redundancy removal for isolated gesture in Indian sign language and recognition using multi-class support vector machine

Sign language is a formal language used by the deaf and dumb people to communicate through bodily movement, especially of hands rather than speech. In this paper, we have presented a vision-based method for recognition of isolated sign considering static and dynamic behaviour of Indian sign language ISL. The proposed methodology consists of three modules: preprocessing, feature extraction and classification. In the preprocessing module, various steps such as skin colour segmentation, redundant frames removal RFR algorithm and face elimination have been performed. The purpose of RFR algorithm is to remove redundant frames from the sign video to speed up the recognition task. In the feature extraction module, multiple features have been extracted. A multi-class support vector machine MSVM and Bayesian K-nearest neighbour BKNN are used to classify the signs. Experimentation with vocabulary of 21 sign from ISL is conducted and the results prove that the proposed method for recognition of gestured sign is effective and having high accuracy. Experimental results demonstrate that the proposed system can recognise signs with 95.3% accuracy.

Zero-stopping constraint-based hybrid tracking model for dynamic and high-dense crowd videos

ABSTRACT Owing to the importance of video surveillance in the public area, tracking finds significant applications using computer vision algorithms to observe the activity of human. In tracking, multi-object tracking is an active research to analyse and detect the activity of anomalies in the crowded scenes. Accordingly, different multi-object tracking algorithms are proposed in the literature to track the human behaviour of the crowded scenes. In this paper, we have presented a zero-stopping criteria-based hybrid tracking algorithm for high-dense crowd videos. Here, head objects are detected using the proposed objective function which considers both colour and texture property of videos. Then, tracking based on motion is performed using the proposed HSIM measure which includes structural similarity (SSIM) and the proposed similarity function. Along with, the data prediction model, exponential weighted moving average (EWMA), is also utilised to track the spatial location of human objects. These two tracking models are then hybridised to obtain the final tracked output. The experimentation is performed with three marathon sequences and the performance is evaluated with particle filtering-based algorithm using tracking number, tracking distance and optimal subpattern assignment metric (OSPA).

Secret image sharing using grayscale payload decomposition and irreversible image steganography

To provide an added security level most of the existing reversible as well as irreversible image steganography schemes emphasize on encrypting the secret image (payload) before embedding it to the cover image. The complexity of encryption for a large payload where the embedding algorithm itself is complex may adversely affect the steganographic system. Schemes that can induce same level of distortion, as any standard encryption technique with lower computational complexity, can improve the performance of stego systems. In this paper, we propose a secure secret image sharing scheme, which bears minimal computational complexity. The proposed scheme, as a replacement for encryption, diversifies the payload into different matrices which are embedded into carrier image (cover image) using bit X-OR operation. A payload is a grayscale image which is divided into frequency matrix, error matrix, and sign matrix. The frequency matrix is scaled down using a mapping algorithm to produce Down Scaled Frequency (DSF) matrix. The DSF matrix, error matrix, and sign matrix are then embedded in different cover images using bit X-OR operation between the bit planes of the matrices and respective cover images. Analysis of the proposed scheme shows that it effectively camouflages the payload with minimum computation time.

A framework for dynamic hand Gesture Recognition using key frames extraction

Hand Gesture Recognition is one of the natural ways of human computer interaction (HCI) which has wide range of technological as well as social applications. A dynamic hand gesture can be characterized by its shape, position and movement. This paper presents a user independent framework for dynamic hand gesture recognition in which a novel algorithm for extraction of key frames is proposed. This algorithm is based on the change in hand shape and position, to find out the most important and distinguishing frames from the video of the hand gesture, using certain parameters and dynamic threshold. For classification, Multiclass Support Vector Machine (MSVM) is used. Experiments using the videos of hand gestures of Indian Sign Language show the effectiveness of the proposed system for various dynamic hand gestures. The use of key frame extraction algorithm speeds up the system by selecting essential frames and therefore eliminating extra computation on redundant frames.

A Novel Framework for Efficient Extraction of Meaningful Key Frames from Surveillance Video

The key frame extraction, aimed at reducing the amount of information from a surveillance video for analysis by human. The key frame is an important frame of a video to provide an overview of the video. Extraction of key frames from surveillance video is of great interest in effective monitoring and later analysis of video. The computational cost of the existing methods of key frame extraction is very high. The proposed method is a framework for Key frame extraction from a long surveillance video with significantly reduced computational cost. The proposed framework incorporates human intelligence in the process of key frame extraction. The results of proposed framework are compared with the results of IMARS IBM multimedia analysis and retrieval system, results of the key frame extraction methods based on entropy difference method, spatial color distribution method and edge histogram descriptor method. The proposed framework has been objectively evaluated by fidelity. The experimental results demonstrate evidence of the effectiveness of the proposed approach.