Automatic Carnatic Music (CM) transcription is an open problem in need of a standardized descriptive notation. The level of detail needed in a descriptive transcription makes it tedious to obtain ground truth by manual means. In this paper, we propose a novel state-based representation of the pitch curve motivated by CM components called constant-pitch notes and stationary points. We also propose a novel transcription technique that uses the Viterbi algorithm to estimate the states and quantized pitch-values. The proposed technique adheres best to raga-notes compared to the existing critical-points technique and uniform quantization. In a listening test, clips synthesized from the proposed notation were rated significantly better (324 ratings, p <; 0.001) than those from critical-points. Further, speed-halving based on state information best matches the actual, observed CM component-duration ratios without losing raga-characteristics. Thus, the proposed transcription can be corrected manually to obtain ground truth and can enhance learning tools.

In this paper, we address the problem of registering a distorted image and a reference image of the same scene by estimating the camera motion that had caused the distortion. We simultaneously detect the regions of changes between the two images. We attend to the coalesced effect of rolling shutter and motion blur that occurs frequently in moving CMOS cameras. We first model a general image formation framework for a 3D scene following a layered approach in the presence of rolling shutter and motion blur. We then develop an algorithm which performs layered registration to detect changes. This algorithm includes an optimisation problem that leverages the sparsity of the camera trajectory in the pose space and the sparsity of changes in the spatial domain. We create a synthetic dataset for change detection in the presence of motion blur and rolling shutter effect covering different types of camera motion for both planar and 3D scenes. We compare our method with existing registration methods and also show several real examples captured with CMOS cameras.

The rule of perspectivity that 'straight-lines-mustremain-straight' is easily inflected in CMOS cameras by distortions introduced by motion. Lines can be rendered as curves due to the row-wise exposure mechanism known as rolling shutter (RS). We solve the problem of correcting distortions arising from handheld cameras due to RS effect from a single image free from motion blur with special relevance to urban scenes. We develop a procedure to extract prominent curves from the RS image since this is essential for deciphering the varying row-wise motion. We pose an optimization problem with line desirability costs based on straightness, angle, and length, to resolve the geometric ambiguities while estimating the camera motion based on a rotation-only model assuming known camera intrinsic matrix. Finally, we rectify the RS image based on the estimated camera trajectory using inverse mapping. We show rectification results for RS images captured using mobile phone cameras. We also compare our single image method against existing video and nonblind RS rectification methods that typically require multiple images.

We present a robust algorithm that registers one point set to another for nonrigid case. We formulate the problem as a Gaussian mixture model (GMM) density estimation by considering one of the point sets as the GMM centroids and the other as the data points generated by GMM. We displace the centroids and make them register to the data by maximizing the likelihood. To facilitate the process, we introduce a set of weights which provide the proximity information among pairs of points of both point sets and iteratively update the displacement and weights in alternating steps. We propose a priority based combination to update the proximity weights, which leverages the richness of Shape Context (SC). In the displacement updation step, we propose a graph-Laplacian regularization which helps in preserving the intrinsic geometry of the point set to be displaced. We also introduce a fast algorithm that reduces the computation complexity significantly. We apply our method on publicly available datasets. Our results validate the robustness of our approach by outperforming current state-of-the-art methods.

Temporal artifacts due to sequential acquisition of measurements in compressed sensing manifest differently from a conventional optical camera. We propose a framework for dynamic scenes to estimate the relative global motion between camera and scene from measurements acquired using a compressed sensing camera. We follow an adaptive block approach where the resolution of the estimated motion path depends on the motion trajectory. To underline the importance of the proposed motion estimation framework, we develop a face recognition algorithm in the compressive sensing domain by factoring in the time-varying nature of the acquisition process.

The authors consider the design of pilot sequences for channel estimation in the presence of carrier frequency offset (CFO) in systems that employ orthogonal frequency division multiplexing (OFDM). The CFO introduces intercarrier interference (ICI) which degrades the accuracy of the channel estimation. In order to minimise this effect, the authors design pilot sequence that minimises the mean square error (MSE) of the modified least squares (mLS) channel estimator. Since the identical pilot sequence, which minimises this MSE, has high peak-to-average power ratio of the OFDM signal, an alternative approach is proposed for channel estimation. The authors first introduce a new estimator as an alternative to the mLS estimator and design a low PAPR pilot sequences tailored to this new estimator. They show that the proposed procedure completely eliminates the effect of the ICI on the channel estimate. They then extend their design of pilot sequences for realistic sparse channels. Both analytical and computer simulation results presented in this study demonstrate the superiority of the proposed approach over conventional methods for channel estimation in the presence of ICI.

Generally face images may be visualized as points drawn on a low-dimensional manifold embedded in high-dimensional ambient space. Many dimensionality reduction techniques have been used to learn this manifold. Orthogonal locality preserving projection (OLPP) is one among them which aims to discover the local structure of the manifold and produces orthogonal basis functions. In this paper, we present a two-step patch based algorithm for face superresolution. In first step a MAP based framework is used to obtain high resolution patch from its low resolution counterpart where the face subspace is learnt using OLPP. To enhance the quality of the image further, we propose a method which uses kernel ridge regression to learn the relation between low and high resolution residual patches. Experimental results show that our approach can reconstruct high quality face images.

Traditional face superresolution methods treat face images as 1D vectors and apply PCA on the set of these 1D vectors to learn the face subspace. Zhang et al [7] proposed Two-directional two-dimensional PCA (2D)2-PCA for efficient face representation and recognition where images are treated as matrices instead of vectors. In this paper, we present a two-step algorithm for face superresolution. In first step, we propose a 2D-framework for face superresolution where the face image is treated as a matrix. (2D)2-PCA is used for learning face subspace and a MAP estimator is used to obtain the global high resolution image from the given low resolution image. To enhance the quality of the image further, we propose a method which uses Kernel Ridge Regression to learn the high frequency component relation between low and high resolution patches of the image. Experimental results show that our approach can reconstruct high quality face images.

Particle filters provide a robust framework for nonlinear and non-Gaussian estimation problems. In this paper, we present a method to incorporate dominant modulation-domain (amplitude modulation-frequency modulation) features into particle filter based template tracking. The dominant AM-FM features capture the local texture structure of an image. Tracking is performed based on correlations of these AM-FM features. We propose an adaptive method that weights these correlations based on the significance of the features for tracking. The performance of the tracker is demonstrated on a few standard test sequences.

We propose a recursive spatial-domain speckle reduction algorithm for synthetic aperture radar (SAR) imagery based on the unscented Kalman filter (UKF) with a discontinuity-adaptive Markov random field (DAMRF) prior. The capability of the UKF in handling speckle noise and the feature preservation ability of the DAMRF model are explored within a unified framework through importance sampling.

This correspondence proposes a recursive algorithm for noise reduction in synthetic aperture radar imagery. Excellent despeckling in conjunction with feature preservation is achieved by incorporating a discontinuity-adaptive Markov random field prior within the unscented Kalman filter framework through importance sampling. The performance of this method is demonstrated on both synthetic and real examples.

This paper presents a novel approach based on the unscented Kalman filter (UKF) for image estimation in film-grain noise. The image prior is modeled as non-Gaussian and is incorporated within the UKF frame work using importance sampling. A small carefully chosen deterministic set of sigma points is used to capture the prior and is propagated through film-grain nonlinearity to compute image statistics. Experimental results are given to demonstrate the efficacy of the proposed method.

This paper presents discontinuity adaptive image estimation within the Kalman filter framework by non-Gaussian modeling of the image prior. A generalized methodology is proposed for specifying state-dynamics using the conditional density of the state given its neighbors, without explicitly defining the state equation. The novelty of our approach lies in directly obtaining the predicted mean and variance of the non-Gaussian state conditional density by importance sampling and incorporating them in the update step of the Kalman filter. Experimental results are given to demonstrate the effectiveness of the proposed method in preserving edges.

Transmission of compressed video over packet networks with nonreliable transport benefits when packet loss resilience is incorporated into the coding. One promising approach to packet loss resilience, particularly for transmission over networks offering dual priorities such as ATM networks, is based on layered coding which uses at least two bitstreams to encode video. The base-layer bitstream, which can be decoded independently to produce a lower quality picture, is transmitted over a high priority channel. The enhancement-layer bitstream(s) contain less information, so that packet losses are more easily tolerated. The MPEG-2 standard provides four methods to produce a layered video bitstream: data partitioning, signal-to-noise ratio scalability, spatial scalability, and temporal scalability. Each was included in the standard in part for motivations other than loss resilience. This paper compares the performance of these techniques (excluding temporal scalability) under various loss rates using realistic length material and discusses their relative merits. Nonlayered MPEG-2 coding gives generally unacceptable video quality for packet loss ratios of 10/sup -3/ for small packet sizes. Better performance can be obtained using layered coding and dual-priority transmission. With data partitioning, cell loss ratios of 10/sup -4/ in the low-priority layer are definitely acceptable, while for SNR scalable encoding, cell loss ratios of 10/sup -3/ are generally invisible. Spatial scalable encoding can provide even better visual quality under packet losses; however, it has a high implementation complexity.

The authors address the problem of adapting the Motion Picture Experts Group (MPEG) quantizer for scenes of different complexity (at bit rates around 1 Mb/s), such that the perceptual quality of the reconstructed video is optimized. Adaptive quantisation techniques conforming to the MPEG syntax can significantly improve the performance of the encoder. The authors concentrate on a one-pass causal scheme to limit the complexity of the encoder. The system employs prestored models for perceptual quality and a bit rate that have been experimentally derived. A framework is provided for determining these models as well as adapting them to locally varying scene characteristics. The variance of an 8*8 (luminance) block is basic to the techniques developed. Following standard practice, it is defined as the average of the square of the deviations of the pixels in the block from the mean pixel value.>

This paper considers the topic of video compression at bit-rates around 1 Mbps for digital storage/playback applications. The backbone algorithm employed is the well-known motion-compensated prediction-error coder. This approach is considerably enhanced by the incorporation of conditional motion-compensated interpolation (CMCI), where some of the video frames are encoded with prediction-error coding and the remainder are encoded with motion-compensated interpolation and interpolation-error coding. The advantage of this combined technique is that interpolation errors can be coded with much fewer bits than prediction errors for similar reconstructed-picture quality. We consider two different techniques for estimating motion in CMCI- coded frames. In order to facilitate direct interaction with the bit-stream, the input sequence is divided into groups such that one frame in each group is intra-coded. The arrangement of predictive- and CMCI-coded frames can be varied; a study of these arrangements is the major focus of this paper. Structures with one-, two- and four-frame interpolation are compared. Our results indicate that one-frame interpolation is inferior to two-frame interpolation on complex test material. Four-frame interpolation is comparable in performance to two-frame interpolation as long as the motion is compensable.

Direct space-domain image coding with Vector Quantization (VQ) has been very effective in the range 0.6 to 1.5 bpp. To achieve quality at lower rates, it is necessary to exploit spatial redundancy over a larger region of pixels than is possible with memoryless VQ. One way to do this is to incorporate memory into a VQ-based coder. A finite-state vector quantizer (FSVQ) employs a state variable which summarizes the past to select one of a family of codebooks to encode each vector. As a result, an FSVQ can achieve the same performance as memoryless VQ at lower rates. In this paper we extend the FSVQ technique to image coding, introducing a novel formulation of the state and state-transition rule using a perceptually-based image classifier. A new distortion measure is proposed for image VQ, which appears to be perceptually more effective than the usual MSE-criterion and leads to a modified LBG design algorithm to obtain codebooks which result in improved edge integrity. At 0.375 bpp, the resulting FSVQ coder achieves performance comparable to earlier memoryless VQs at 0.7 bpp.

Vector Quantization is an attractive block coding scheme which allows more efficient use of a given channel capacity. The computational complexity of the codebook search, however, limits the practical applications of Vector Quantization -- especially in real-time situations. This paper presents a special purpose processing architecture which supports real-time search of large codebooks. The core of the computation required for a squared-error distortion measure is accomplished using a VLSI pattern matching chip. We describe a pipeline architecture which reduces the effective computation time for a single vector component to one period. The chip supports a selectable vector dimension of 2, 4, 8, or 8 and can be used with any codebook size consistent with the input vector rate and the chip's throughput. We have built and tested a 4-micron NMOS chip which supports up to four million squared-error distance calculations per second. We outline three approaches to real-time Vector quantization which use this chip to do the pattern matching: two waveform coding processors using Vector Pulse Code Modulation and Adaptive Vector Predictive Coding, and a novel approach to rapid codebook design.