Tanushyam Chattopadhyay

A Survey on Different Video Watermarking Techniques and Comparative Analysis with Reference to H.264/AVC

Last few years have witnessed rapid growth in video coding technology. Among various standards, H.264/advanced video codec (AVC) is found to be of significant importance regarding reduced bandwidth, better image quality and network friendliness. One of the current fields of interest is to develop a system with authentication and copyright protection methodology embedded within an efficient video codec. In this paper we first perform a survey on available video watermarking techniques, feasibility study on watermarking techniques meeting application specific criteria for H.264/AVC and then we perform a comparative analysis based on robustness and computational complexity of different watermarking algorithms

Human Activity Recognition from Kinect Captured Data Using Stick Model

In this paper authors have presented a method to recognize basic human activities such as sitting, walking, laying, and standing in real time using simple features to accomplish a bigger goal of developing an elderly people health monitoring system using Kinect. We have used the skeleton joint positions obtained from the software development kit (SDK) of Microsoft as the input for the system. We have evaluated our proposed system against our own data set as well as on a subset of the MSR 3Ddaily activity data set and observed that our proposed method out performs state-of-the-art methods.

Automatic Selection of Binarization Method for Robust OCR

Many algorithms are now available for doing the same task (e.g. binarization, page segmentation, character recognition, etc.) in document image analysis (DIA) and choosing a particular algorithm(s) for a particular task is often a non-trivial problem. This paper proposes a model for automatically selecting the correct algorithm(s) for a given problem. Binarization has been taken a reference to illustrate the proposed approach. Several previously unexplored issues are addressed in this work. For example, only one method may not be good for the binarization of an entire document whereas a particular method may produce desired result for a particular region. Therefore, for a given document image, our model selects a set of one or more binarization techniques suitable for different regions of the document. This selection is completely automatic and guided by the machine learning approaches. Formulation of a completely automatic way for generating the annotated data for training the learning algorithms is also a novel contribution of this work. Evaluation of the approach is done using ICDAR 2003 Robust Reading data set and results highlight the potential of the proposed approach for automatic selection of correct DIA algorithm(s) from a set of several alternatives.

Generation of electronic program guide for RF fed TV channels by recognizing the channel logo using fuzzy multifactor analysis

This paper presents a novel method for providing the Electronic Program Guide (EPG) for the Radio Frequency (RF) fed cable TV users by recognizing the channel logo and then extracting the EPG from their website. The prototype is developed in X86 platform and then ported on a real time DSP platform and performs with nearly 100% accuracy.

Recognition of trademarks from sports videos for channel hyperlinking in consumer end

In this paper authors have proposed a system to automatically recognize the Trademarks from sports video for channel hyperlinking in client end. In this method we have used the output of Set Top Box (STB) video stream in YUV 4:2:2 formats as input to our application. In this work we have first localized the text regions using some characteristic of text and then recognized the trademark using the shape invariant features and color features from the restricted trademark database. Experimental results show that the proposed approach can work in real time in any commercially available DSP platform and can mark the trademarks in the video successfully. The system on different type of sports videos gives a recall rate of 86.6% and a precision rate of 85.42%.

Action recognition using joint coordinates of 3D skeleton data

We propose an action recognition technique using the 3D skeleton model of human without compromising the identity of the person. The skeleton model is defined as a set of 3D joint (e.g. knee or hip joint) coordinates obtained from the Kinect. The low frequency sensor noise in estimating the joint coordinates is removed after modeling the covariance matrix of the joint coordinates as a function of variance of individual joint coordinates. We determine a range for the threshold of this covariance matrix to detect active joints defining an action. Since, a sparse set of active joint coordinates is enough to represent an action, we map these coordinates to lower dimensional linear manifold before training using an SVM classifier. The recognition rate using our proposed approach outperforms competing approaches by at least 2%.

Adaptive rate control for H.264 based video conferencing over a low bandwidth wired and wireless channel

This paper presents a method for adaptive rate control in a video conferencing solution over a variable bandwidth channel. The bandwidth is estimated using the round-trip delay of probe packets. Based on the estimated bandwidth, the video rate adaptation is done in two folds; one in the H.264 video encoder based on adaptive basic unit selection and the other by generation of fragments for encoding data and controlling the transmission delay of the same. The audio adaptation is done with the voice activity detection (VAD) of NB-AMR speech codec. System architecture is proposed for a generic video conferencing solution to implement the multimode adaptive rate control for a variable Quality of service (QoS) wired channel and a low bandwidth wireless channel. The implementation is tested with ADSL and CDMA 1xRTT channels. The proposed method gives an improvement in image quality (PSNR) compared to the reference H.264 (JM9.5) encoder demonstrating an improved adaptive rate control behavior in the heterogeneous network.

A Weighted Finite-State Transducer (WFST)-Based Language Model for Online Indic Script Handwriting Recognition

Though designing of classifies for Indic script handwriting recognition has been researched with enough attention, use of language model has so far received little exposure. This paper attempts to develop a weighted finite-state transducer (WFST) based language model for improving the current recognition accuracy. Both the recognition hypothesis (i.e. the segmentation lattice) and the lexicon are modeled as two WFSTs. Concatenation of these two FSTs accept a valid word(s) which is (are) present in the recognition lattice. A third FST called error FST is also introduced to retrieve certain words which were missing in the previous concatenation operation. The proposed model has been tested for online Bangla handwriting recognition though the underlying principle can equally be applied for recognition of offline or printed words. Experiment on a part of ISI-Bangla handwriting database shows that while the present classifiers (without using any language model) can recognize about 73% word, use of recognition and lexicon FSTs improve this result by about 9% giving an average word-level accuracy of 82%. Introduction of error FST further improves this accuracy to 93%. This remarkable improvement in word recognition accuracy by using FST-based language model would serve as a significant revelation for the research in handwriting recognition, in general and Indic script handwriting recognition, in particular.

Recognition of channel logos from streamed videos for value added services in connected TV

This paper presents a novel method for recognizing the channel logos from the streamed videos in real time, which has various applications for value added services in the connected TV space. The results presented are based on the accuracy and performance in terms of time complexity of the channel logo recognition algorithm. The prototype is developed in X86 platform and then ported on a commercially available DSP with nearly 100% accuracy in real time.

Enhancements of H.264 Encoder performance for video conferencing and videophone applications in TMS320C55X

In this age of multimedia convergence, video based applications for conferencing and streaming are seeing big market traction. The main challenge of implementing these systems is a tradeoff between several factors like bandwidth, image quality, implementation cost and speed (in terms of mega cycles per second). Moreover to implement this encoder-decoder system (CODECs) in a real time system based on any digital signal processor (DSP) is more difficult because of their restricted resources like memory and CPU speed. In this paper, we present some novel approach, which reduces the computational complexity and also meets the memory constraint of the target processor for a standard H.264 baseline encoder without sacrificing the rate-distortion performance. The proposed algorithms are applied on standard test sequence of different resolutions. The results obtained from a considerably large number of test sequences show the strength of our proposed algorithm. We claim the betterment in performance measured in mega cycles per second (MCPS) at the cost of negligible loss in image quality. Moreover we gain the betterment in image quality using algorithm modification in rate controlling. Implementation details are also presented for a QCIF H.264 baseline encoder in a TMS320C55x DSP, which has only 256 KB RAM and 150 MHz clock speed. The enhanced performance is achieved thorough formulation of novel adaptive algorithms for rate control and motion estimation. We achieve almost 40% MCPS reduction at the cost of less than 1% reduction in image quality. We also get betterment in image quality by our proposed rate-controlling algorithm