R. Venkatesh Babu

Robust tracking with motion estimation and local Kernel-based color modeling

Visual tracking has been a challenging problem in computer vision over the decades. The applications of visual tracking are far-reaching, ranging from surveillance and monitoring to smart rooms. Mean-shift tracker, which gained attention recently, is known for tracking objects in a cluttered environment. In this work, we propose a new method to track objects by combining two well-known trackers, sum-of-squared differences (SSD) and color-based mean-shift (MS) tracker. In the proposed combination, the two trackers complement each other by overcoming their respective disadvantages. The rapid model change in SSD tracker is overcome by the MS tracker module, while the inability of MS tracker to handle large displacements is circumvented by the SSD module. The performance of the combined tracker is illustrated to be better than those of the individual trackers, for tracking fast-moving objects. Since the MS tracker relies on global object parameters such as color, the performance of the tracker degrades when the object undergoes partial occlusion. To avoid adverse effects of the global model, we use MS tracker to track local object properties instead of the global ones. Further, likelihood ratio weighting is used for the SSD tracker to avoid drift during partial occlusion and to update the MS tracking modules. The proposed tracker outperforms the traditional MS tracker as illustrated.

Recognition of human actions using motion history information extracted from the compressed video

Human motion analysis is a recent topic of interest among the computer vision and video processing community. Research in this area is motivated by its wide range of applications such as surveillance and monitoring systems. In this paper we describe a system for recognition of various human actions from compressed video based on motion history information. We introduce the notion of quantifying the motion involved, through what we call Motion Flow History (MFH). The encoded motion information readily available in the compressed MPEG stream is used to construct the coarse Motion History Image (MHI) and the corresponding MFH. The features extracted from the static MHI and MFH compactly characterize the spatio-temporal and motion vector information of the action. Since the features are extracted from the partially decoded sparse motion data, the computational load is minimized to a great extent. The extracted features are used to train the KNN, Neural network, SVM and the Bayes classifiers for recognizing a set of seven human actions. The performance of each feature set with respect to various classifiers are analyzed. q 2003 Elsevier B.V. All rights reserved.

Robust object tracking with background-weighted local kernels

Object tracking is critical to visual surveillance, activity analysis and event/gesture recognition. The major issues to be addressed in visual tracking are illumination changes, occlusion, appearance and scale variations. In this paper, we propose a weighted fragment based approach that tackles partial occlusion. The weights are derived from the difference between the fragment and background colors. Further, a fast and yet stable model updation method is described. We also demonstrate how edge information can be merged into the mean shift framework without having to use a joint histogram. This is used for tracking objects of varying sizes. Ideas presented here are computationally simple enough to be executed in real-time and can be directly extended to a multiple object tracking system.

No-reference JPEG-image quality assessment using GAP-RBF

In this paper, we present a novel no-reference (NR) method to assess the quality of JPEG-coded images using a sequential learning algorithm for growing and pruning radial basis function (GAP-RBF) network. The features for predicting the perceived image quality are extracted by considering key human visual sensitivity factors such as edge amplitude, edge length, background activity and background luminance. Image quality estimation involves computation of functional relationship between HVS features and subjective test scores. Here, the functional relationship is approximated using GAP-RBF network. The advantage of using sequential learning algorithm is its capability to learn new samples without affecting the past learning. Further, the sequential learning algorithm requires minimal memory and computational effort. Experimental results prove that the prediction of the trained GAP-RBF network does emulate the mean opinion score (MOS). The subjective test results of the proposed metric are compared with JPEG no-reference image quality index as well as full-reference structural similarity image quality index and it is observed to outperform both.

DeepFix: A Fully Convolutional Neural Network for Predicting Human Eye Fixations

Understanding and predicting the human visual attention mechanism is an active area of research in the fields of neuroscience and computer vision. In this paper, we propose DeepFix, a fully convolutional neural network, which models the bottom–up mechanism of visual attention via saliency prediction. Unlike classical works, which characterize the saliency map using various hand-crafted features, our model automatically learns features in a hierarchical fashion and predicts the saliency map in an end-to-end manner. DeepFix is designed to capture semantics at multiple scales while taking global context into account, by using network layers with very large receptive fields. Generally, fully convolutional nets are spatially invariant—this prevents them from modeling location-dependent patterns (e.g., centre-bias). Our network handles this by incorporating a novel location-biased convolutional layer. We evaluate our model on multiple challenging saliency data sets and show that it achieves the state-of-the-art results.

SeamSeg: Video Object Segmentation Using Patch Seams

In this paper, we propose a technique for video object segmentation using patch seams across frames. Typically, seams, which are connected paths of low energy, are utilised for retargeting, where the primary aim is to reduce the image size while preserving the salient image contents. Here, we adapt the formulation of seams for temporal label propagation. The energy function associated with the proposed video seams provides temporal linking of patches across frames, to accurately segment the object. The proposed energy function takes into account the similarity of patches along the seam, temporal consistency of motion and spatial coherency of seams. Label propagation is achieved with high fidelity in the critical boundary regions, utilising the proposed patch seams. To achieve this without additional overheads, we curtail the error propagation by formulating boundary regions as rough-sets. The proposed approach out-perform state-of-the-art supervised and unsupervised algorithms, on benchmark datasets.

Data-free Parameter Pruning for Deep Neural Networks.

Deep Neural nets (NNs) with millions of parameters are at the heart of many state-of-the-art computer vision systems today. However, recent works have shown that much smaller models can achieve similar levels of performance. In this work, we address the problem of pruning parameters in a trained NN model. Instead of removing individual weights one at a time as done in previous works, we remove one neuron at a time. We show how similar neurons are redundant, and propose a systematic way to remove them. Our experiments in pruning the densely connected layers show that we can remove upto 85\% of the total parameters in an MNIST-trained network, and about 35\% for AlexNet without significantly affecting performance. Our method can be applied on top of most networks with a fully connected layer to give a smaller network.

Compressed domain action classification using HMM

Abstract This paper proposes three techniques of feature extraction for person independent action classification in compressed MPEG video. The features used are extracted from motion vectors, obtained by partial decoding of the MPEG video. The feature vectors are fed to Hidden Markov Model (HMM) for classification of actions. Totally seven actions were trained with distinct HMM for classification. Recognition results of more than 90% have been achieved. This work is significant in the context of emerging MPEG-7 standard for video indexing and retrieval.

Human action recognition using depth maps

In this paper we propose an approach to recognize human actions using depth images. Here, we capture the motion dynamics of the object from the depth difference image and average depth image. The features from the space-time depth difference images are obtained from hierarchical division of the silhouette bounding box. We also make use of motion history images to represent the temporal information about the action. We make use of the translation, scale and orientation invariant Hu moments to represent the features of the motion history image and the average depth image. We then classify human actions using support vector machines. We analyze the representation efficiency of Hu moments and the hierarchical division of bounding boxes separately in order to evaluate the contribution of each of the features. The results show superior performance of over 90% when both features are combined.

Compressed domain motion segmentation for video object extraction

This paper addresses the problem of extracting video objects from MPEG compressed video. The only cues used for object segmentation are the motion vectors which are sparse in MPEG. A method for automatically estimating the number of objects and extracting independently moving video objects using motion vectors is presented here. First, the motion vectors are accumulated over few frames to enhance the motion information, which are further spatially interpolated to get a dense motion vectors. The final segmentation from the dense motion vectors is obtained by applying Expectation Maximization (EM) algorithm. A block based affine clustering method is proposed for determining the number of appropriate motion models to be used for the EM step. Finally, the segmented objects are temporally tracked to obtain the video objects. This work has been carried out in the context of the emerging MPEG-4 standard which aims at interactivity at object level.