Detection of small, point targets is fundamental in applications such as early warning systems, surveillance, astronomy, and microscopy. The presence of noise and clutter can make it challenging to detect small targets while minimizing false detections. This paper presents a method for infrared small target detection using convolutional neural networks. The proposed method augments a conventional space-based detection processing chain with a lightweight neural network to predict the probability that a detection is a target. The proposed network is trained on 7 × 7 pixel windows using both the image sequence and the respective background-subtracted images. Results show that our method improves probability of detection at low false detection rates.

Automatic building segmentation is an important task for satellite imagery analysis and scene understanding. Most existing segmentation methods focus on the case where the images are taken from directly overhead (i.e., low off-nadir/viewing angle). These methods often fail to provide accurate results on satellite images with larger off-nadir angles due to the higher noise level and lower spatial resolution. In this paper, we propose a method that is able to provide accurate building segmentation for satellite imagery captured from a large range of off-nadir angles. Based on Bayesian deep learning, we explicitly design our method to learn the data noise via aleatoric and epistemic uncertainty modeling. Satellite image metadata (e.g., off-nadir angle and ground sample distance) is also used in our model to further improve the result. We show that with uncertainty modeling and metadata injection, our method achieves better performance than the baseline method, especially for noisy images taken from large off-nadir angles1.

When a disaster strikes, accurate situational information and a fast, effective response are critical to save lives. High resolution satellite images enable emergency responders to estimate the location, cause, and severity of damage. In this paper, we present a Siam-U-Net-Attn model with an attention mechanism to assess the damage level of buildings given a pair of satellite images showing a scene before and after a disaster. We evaluate the proposed method on xView2, a building damage assessment dataset, and demonstrate that the proposed approach achieves accurate damage scale classification and building segmentation.

Retinal vessel extraction and segmentation is essential for supporting diagnosis of eye-related diseases. In recent years, deep learning has been applied to vessel segmentation and achieved excellent performance. However, these supervised methods require accurate hand-labeled training data, which may not be available. In this paper, we propose an unsupervised segmentation method based on our previous connected tube marked point process (MPP) model. The vessel network is extracted by the connected-tube MPP model first. Then a new tube-based segmentation method is applied to the extracted tubes. We test this method on STARE and DRIVE databases and the results show that not only do we extract the retina vessel network accurately, but we also achieve high G-means score for vessel segmentation, without using labeled training data.

Road detection in high-resolution satellite images is an important and popular research topic in the field of image processing. In this paper, we propose a novel road extraction and tracking method based on road segmentation results from a convolutional network, providing improved road detection. The proposed method incorporates our previously proposed connected-tube marked point process (MPP) model and a post-tracking algorithm. We present experimental results on the Massachusetts roads dataset to show the performance of our method on road detection in remotely-sensed images.

Differentiating objects separated by narrow gaps is a challenging and important task in analyzing microscopic images. These small separations provide useful information for applications that require detailed boundary information and/or an accurate particle count. We present a new approach to the modeling of these gaps based on a marked point process(MPP) framework. We propose to model narrow gaps as geometric structures called channels, and define Gibbs energies for these models. The Reversible-jump Markov chain Monte Carlo(RJMCMC) algorithm embedded with simulated annealing is used as an optimization method, and the switching kernel in an RJMCMC is newly designed to speed up the algorithm. In this paper, we also propose a method to exploit a detected channel configuration to reduce bridging channel defects in conventional segmentation algorithms. Experimental results demonstrate that the proposed channel modeling methods are successful in detecting gaps between closely adjacent objects. The results also show that the proposed interaction parameter control method improves boundary precision in the segmentation of microscopic images.

In this paper, we propose a connected-tube model based on a Marked Point Process (MPP) for strip feature extraction in images. This model incorporates a connection prior that favors certain connections between tubes based on their mutual positional relationship. Moreover, this model can easily be combined with other geometric models to form a mixed MPP model for more complex detection tasks. The proposed tube model is applied to fiber detection in microscopy images by combining connected-tube and ellipse models. The ellipse model is used for detecting short fibers, while the longer fibers are detected by tube model. We also test the model on road and building detection in remotely sensed images.

Stochastic models of images are very useful for applications such as segmentation, deblurring, and reconstruction. Sometimes it is important to be able to simulate, or draw samples from, a stochastic image model. For example, simulation can be used as an optimization tool for segmenting, deblurring, or reconstructing an image. Also, simulation of images that characterize a system can be helpful in understanding the system, by allowing virtual exploration of models of the system instead of expensive and time-intensive physical experimentation. There are of course many Markov chain Monte Carlo (MCMC) methods for drawing samples from the ubiquitous Markov random field (MRF) image model. However, these methods draw sample images that represent typical cases of the model. To sample images that occur with low probability, which represent rare events, a prohibitive number of Monte Carlo samples would need to be drawn using traditional MCMC. In this paper, we turn to large deviations theory and importance sampling to propose a rare-event simulation method for MRFs. We then use an impactful problem from materials science to demonstrate the application of our method. More specifically, we look at the phenomenon of abnormal grain growth in polycrystalline materials. With our proposed method, we consistently generate images containing abnormal grain growth, though this is a very challenging problem for standard Monte Carlo simulation methods. Importantly, our method can be used to simulate rare events in a broad class of imaging applications, namely those that use an MRF model.

In this paper, using the Marked Point Process (MPP) framework, we propose a new model and method to detect channels in microscope images of materials. The detected channel configuration information is expected to be useful for analyzing materials images in many ways. As one example of the possible usage of this channel information, we propose a method to reduce bridging channel defects in the segmentation of images containing channels. Our experimental results demonstrate that proposed model and methods works effectively for both detecting channels and aiding segmentation.

Tunable Diode Laser Absorption Spectroscopy (TDLAS) is an emerging technique for simultaneous sensing of temperature and concentration of gaseous media. However, simultaneous reconstruction of temperature and concentration using TDLAS measurements is a nonlinear inverse problem and unlike other forms of computed tomography (CT), it is typically not possible to take a large number of projection measurements; so reconstructions are often computed using simplistic assumptions that limit the usability of the results. In this paper, we present a fast algorithm for model-based iterative reconstruction (MBIR) of TDLAS data. Our TDLAS-MBIR method uses a nonlinear forward model based on the physics of light absorption and incorporates a holistic prior model that can be learned from very sparse training data. Reconstructions performed on computational fluid dynamics (CFD) phantoms show that our proposed reconstruction algorithm is fast; works well when the number of pixels, p, far exceeds the number of measurements, M; is robust against noise; and produces good reconstructions using few training examples for the prior model.

The Markov Random Field (MRF) has been used extensively in Image Processing as a means of smoothing interfaces between differing regions in an image. The MRF applies a total boundary length `energy' penalty that is subsequently minimized by an inversion algorithm. The minimization of energy implies a force associated with boundaries, the sum of which must equal zero at every point at equilibrium. This requirement leads to long range interactions, resulting from the short-range interactions of the MRF, which biases segmentation results. This work uses a simple Bayesian MRF regularized segmentation method to show that classical results from Surface Science are reproduced when segmenting regions of low contrast. This has implications, both in the Materials Science and Image Processing fields.

Markov random field (MRF) models have become extremely popular for regularization in image segmentation. They are useful for imposing local constraints, but they have limited capability for imposing global constraints. Marked Point Process (MPP) models incorporate global information, such as shape, as a prior, but local constraints, such as pixel-wise interaction, are not easily modeled. In this paper, we propose a combined MRF and MPP model and demonstrate its usefulness for micrograph analysis. Our hybrid model imposes both local and global constraints, at the pixel level as well as the object level. A multiple birth and death algorithm is then used to approximate the MAP segmentation using the hybrid model. We present experimental results to show the advantage of our model over the MPP for object identification and the MRF for segmentation.

Multiple Description Coding (MDC) is an error resilient video coding method suitable for real-time applications when retransmission is unacceptable. A four description MDC architecture is described with temporal and spatial error concealment capabilities at the decoder. In this paper, we present an adaptive error concealment method using error estimation. Experimental results demonstrate the efficacy of the proposed method.

Multiple Description Coding (MDC) is an efficient error resilient video coding method especially for real-time applications when retransmission is unacceptable. For applications with scalable, multicast and P2P environments, it is advantageous to use more than two descriptions. In this paper, we present an adaptive temporal-spatial error concealment method using motion vector analysis to improve the performance of MDC. Experimental results demonstrate the efficacy of the proposed method.

Multiple Description Coding (MDC) is one of the most efficient error resilient video coding methods especially when retransmissions are not acceptable. For applications in scalable, multicast and P2P environments, it is advantageous to use more than two descriptions. In this paper, we present two macroblock-level adaptive error concealment methods to improve the performance of MDC with four descriptions. A Gilbert model is used as the channel model to simulate burst packet loss. Experimental results demonstrate the efficacy of the two proposed methods.

We propose the use of large tree-structured dictionaries of tilings for video compression. Our first contribution is the construction of a rate-distortion cost function that admits fast search algorithms to select the optimal tiling for the motion compensation stage of a video coder. The computation of the cost is enabled through novel algorithms to approximate the bit rate and the distortion. Our second contribution is an efficient arithmetic coding algorithm to encode the selected tree-structured tiling. We illustrate the effectiveness of our approach by showing that a H.264/AVC-like video coder utilizing one of the proposed tiling selection methods results in up to 16% savings in bit rate for several standard video sequences as compared to H.264/AVC. This is accomplished with only a modest increase in the computation time at the encoder.

A primary challenge in multi casting video in a wireless LAN to multiple clients is to deal with the client diversity -- clients may have different channel characteristics and hence receive different numbers of transmissions from the AP. A promising approach to overcome this problem is to combine multi-resolution (layered) video coding with interlayer network coding. The fundamental challenge in such an approach is to determine the strategy of coding the packets across different layers that maximizes the number of decoded layers at all clients. This paper makes three contributions. (1) We first show that even for one client, the previously proposed canonical triangular scheme for inter-layer network coding can perform poorly. We show how to enhance the triangular scheme by incorporating the estimated target number of layers which significantly improves its effectiveness. (2) We show that such an enhanced triangular scheme still performs poorly for multiple clients with diverse channel characteristics, which motivates the need for searching for the optimal coding strategy. The naive way of searching for the optimal strategy is computationally prohibitive. We present several optimizations that drastically reduce the complexity of exhaustively searching for the optimal strategy, making it feasible in real time. (3) Finally, we design and evaluate an on line video delivery scheme, Percy, to be deployed at a proxy behind the AP of a wireless LAN. Our simulation results show that Percy outperforms the previous inter-layer coding heuristic by up to 22-80% with varying numbers of clients.

Multiple Description Coding (MDC) is one of the most efficient methods to combat error-prone channels especially when retransmission is unacceptable. In applications involving scalable, multicast and P2P environments, it is advantageous to use more than two descriptions. In this paper, we present an adaptable temporal-spatial error concealment method based on error tracking to improve the performance of MDC. A Gilbert model is used as the channel model for burst packet loss simulation. Experimental results demonstrate the efficacy of the proposed method.

A primary challenge in multicasting video in a wireless LAN is to deal with the client diversity -- clients may have different channel characteristics and hence receive different numbers of transmissions from the AP. A promising approach to overcome this problem is to combine scalable video coding techniques such as MRC or MDC, which divide a video stream into multiple substreams, with inter-layer network coding. The fundamental challenge in such an approach is to determine the strategy of coding the packets across different layers that maximizes the number of decoded layers at all clients. In [7], the authors showed that inter-layer NC indeed helps the delivery of MRC coded media over the WiFi, and proposed how to efficiently search for the optimal coding strategies online.

We present in this paper the rate distortion performance analysis of the pyramid motion compensation for video spatial scalability. Theoretical performance function is derived in closed form based on quantization noise modeling. Evaluation of the rate distortion performance functions shows that, compared to intra-layer motion-compensated prediction coding the enhancement layer, pyramid method is expected to achieve more efficient compression if the base layer is encoded at a sufficient high quality, i.e., the encoded rate is high enough.

We propose a new method for segmenting noisy image sequences in a way that imposes consistency between neighboring segmentations in the sequence. Our method uses a statistical model composed of a spatial Markov random field model and a temporal Markov chain model. Results from segmenting sequences of microscopy images of growing silicon nanowires using the proposed model and method show significant improvement over segmenting the sequences using 2D segmentation.

One of the most difficult problems in video transmission is communication over error-prone channels, especially when retransmission is unacceptable. To address this problem, Multiple Description Coding (MDC) has been proposed as an effective solution due to its robust error resilience. Considering applications in scalable, multicast and P2P environments, it is advantageous to use more than two descriptions (which is designated multi-description MDC in this paper). In this paper, we present a new four-description MDC for high lossrate channel using a hybrid structure of temporal and spatial correlations. A Gilbert model is used as the channel model for burst packet loss simulation. Experimental results demonstrate the efficacy of the proposed method.

We present in this letter the rate distortion (RD) performance analysis of the inter-layer subband and pyramid motion compensation techniques for spatially scalable video coding. Theoretical performance functions are derived based on RD theory and quantization noise modeling. We assume the base layer is encoded by a non-scalable coder. The coding efficiency of the enhancement (enh.) layer, measured by the signal-to-noise ratio, is determined by the input video power spectral density, the motion estimation error distribution, and the base layer encoder rate. Numerical evaluations of the performance functions show that, compared to independent motion-compensated prediction encoding of the enh. layer, the inter-layer pyramid and subband methods are expected to be more efficient if the base layer is encoded at a sufficiently high quality or the motion estimation accuracy is relatively low in the enh. layer. Results from real video data encoding show that the presented theoretical analysis can be very useful to understand the efficiencies of these spatial scalability techniques.

In this paper, we derive the rate distortion lower bounds of spatially scalable video coding techniques. The methods we evaluate are subband and pyramid motion compensation where temporal redundancies in the same spatial layer as well as interlayer spatial redundancies are exploited in the enhancement layer encoding. The rate distortion bounds are derived from rate distortion theory for stationary Gaussian signals where mean square error is used as the distortion criteria. Assuming that the base layer is encoded by a non-scalable video coder, we derive the rate distortion functions for the enhancement layer, which depend on the power spectral density of the input signal, the motion prediction error probability density function and the base layer encoding performance. We will show that pyramid and subband methods are expected to outperform independently encoding the enhancement layer using motion-compensated prediction, in terms of rate distortion efficiency, when the base layer is encoded at a relatively higher quality or less accurate displacement estimation happens in the enhancement layer.

In this paper, we derive the rate distortion performance lower bound of the subband motion compensation technique which exploits both the inter-layer spatial and intra-layer temporal redundancies for video spatial scalability. The lower bound is derived by rate distortion theory for stationary Gaussian signals where mean square error (MSE) is used as the distortion criteria. Assuming that the base layer is encoded by an optimal non-scalable video coder, the rate distortion function derived for the enhancement layer depends on only the power spectral density (PSD) of the input signal, the motion prediction error probability density function (p.d.f.) and the base layer encoding parameter. Our evaluations will show that, if base layer is encoded at a relatively high quality, subband method is expected to outperform simulcast, where spatial layers are encoded independently, in terms of rate distortion efficiency.

We investigate the automated segmentation of microstructures of a nickel-based superalloy using digital microscopy data. We study the combination of a region merging segmentation method called the stabilized inverse diffusion equation (SIDE), and a stochastic segmentation method, the expectation-maximization/maximization of the posterior marginals (EM/MPM) algorithm. We use the SIDE algorithm to segment the grain boundaries and we use the EM/MPM algorithm to classify two phases of the material within each grain. Experimental results demonstrate the effectiveness of our approach.

This paper presents an adaptive motion compensation technique for encoding the enhancement layers in spatially scalable video coding. Based on the base layer reconstructed quantized prediction error, quantization parameters and motion vectors of both the base and enhancement layers, the proposed approach automatically selects either the inter-layer subband motion compensation, or encoding the enhancement layer without using any information from the base layer, which we call independent encoding. Compared to the SVC spatial scalability, the proposed adaptive method achieves an improvement up to about 0.6 dB with about 33% encoder computational complexity reduction.

In this paper, an efficient inter-layer motion-compensation technique is proposed for enhancement layer in spatially scalable video coding. The proposed approach exploits the prediction residue correlation between consecutive spatial layers. In enhancement-layer motion compensation, it selects the method with the lowest rate-distortion (RD) cost from three schemes: pyramid motion compensation, subband motion compensation, and encoding the enhancement layer with no base-layer information. To reduce the block artifacts of the enhancement-layer prediction used in the subband method, extended edge prediction is introduced in the proposed approach. Compared with the current scalable H.264/AVC extension (SVC) motion-compensation scheme, our experimental results show that the proposed method exploits the redundancy between the base layer and the enhancement layer more efficiently and hence improves the coding performance. It achieves an improvement up to about 1dB in the enhancement-layer encoding for different coding parameters and video sequences.

It has been recently discovered that a faithful representation for the shape of some simple distributions can be constructed using invariant statistics [1,2]. In this paper, we consider the more general case of a Gaussian mixture model. We show that the shape of generic Gaussian mixtures can be represented without any loss by the distribution of the distance between two points independently drawn from this mixture. In other words, we show that if their respective distributions of distances are the same, then there exists a rigid transformation mapping one Gaussian mixture onto the other. Our main motivation is the problem of recognizing the shape of an object represented by points given noisy measurements of these points which can be modeled as a Gaussian mixture.

This paper describes an efficient inter-layer motion compensation approach for enhancement layer prediction in spatially scalable video coding. The proposed approach provides improved performance compared to the current motion compensation scheme in the SVC standard. The improvement in performance is achieved by adding a new mode for enhancement layer motion compensation. Our experimental results show that the proposed method achieves an improvement up to about 1 dB in a two-layer spatially scalable framework compared to current SVC enhancement layer prediction.

In this paper, a new enhancement layer motion compensation technique referred to as subband motion compensation is proposed for spatially scalable video coding. This approach is proposed as an alternative to a technique which we call pyramid motion compensation in this paper and which is referred to as inter-layer residual prediction in the H.264/MPEG4-AVC scalable extension, Scalable Video Coding (SVC) standard. The main difference between these two techniques lies in the way they use the base layer information to encode the enhancement layer. Experimental results comparing the two approaches show that for enhancement layer encoding, pyramid method is better when the corresponding base layer is encoded with a lower bitrate while subband method outperforms pyramid method when the base layer has a higher bitrate. This motivates future proposed work to adaptively choose between these two methods at the macroblock level or even at the transform coefficient level for spatially scalable video coding.

We build on our previous work on best basis selection algorithms in large tree-structured dictionaries, previously used to construct effective image coders based on block and lapped transforms. In the present paper, we use such algorithms to select the optimal tiling for the motion compensation step in video compression. We illustrate the effectiveness of this approach by showing that our tiling selection method results in up to 18% savings in bit rate as compared to the H.264 tiling selection, for several standard video sequences.

This paper describes a new technique for reducing the blockiness that results from using the reduced-resolution update (RRU) video coding tool. In RRU coding, full resolution prediction error residuals are downsampled and coded at a reduced spatial resolution. In previous versions of RRU, such as that defined in Annex Q of the H.263 standard, each block of prediction errors is processed without using samples from any neighboring blocks. This can lead to severe blockiness in the decoded picture. To reduce this blockiness H.263 Annex Q RRU increases the deblocking filter strength to the maximum value. In the new technique presented here, referred to as RRU+, the downsampling and interpolation filters use residuals from neighboring blocks to prevent the blockiness that results from RRU. Experimental results comparing RRU+, RRU with normal deblocking filter strength (referred to simply as RRU in this paper), and RRU with maximum deblocking filter strength are presented. These results show that, while RRU+ and RRU with maximum deblocking strength both eliminate the blocking artifacts associated with RRU, RRU+ provides better peak signal-to-noise ratio (PSNR) performance than RRU, whereas RRU with maximum deblocking strength degrades the PSNR performance compared to RRU. Results comparing H.264 with RRU/RRU+ are also presented and the effectiveness of RRU/RRU+ as a tool for improving coding efficiency is discussed.

This paper describes a new improvement upon a video coding tool known as reduced-resolution update (RRU). In RRU, as described in Annex Q of the H.263 standard, full resolution prediction error residuals are downsampled and coded at a reduced spatial resolution. During RRU coding, as previously proposed, each block of prediction errors is downsampled and interpolated without reference to any of its neighboring blocks. This can lead to severe blockiness in the decoded picture. In H.263 Annex Q RRU, the deblocking filter strength is increased to the maximum value to reduce this blockiness. In the new technique presented here, referred to as RRU+, the downsampling and interpolation filters use residuals from neighboring blocks to prevent the blockiness that results from RRU. Experimental results show that RRU+ provides better PSNR versus bitrate performance than RRU with maximum deblocking filter strength.

A new motion compensation (MC) scheme for fine-grain scalable video coding is proposed. This new scheme uses two MC loops, one for the base layer and one for the enhancement layer. A technique called conditional replacement (CR), which adaptively selects between the base layer and enhancement layer prediction for each enhancement layer DCT coefficient, is used to simultaneously improve coding efficiency and reduce prediction drift, as compared to a scheme that always uses the enhancement layer prediction. A new CR architecture that. uses reference frames stored in the spatial domain, unlike the DCT-domain CR architecture previously proposed for frequency scalability, is presented. Whereas a straightforward implementation of a "spatial-domain" CR would require two extra DCTs for each block, the new architecture uses only one extra DCT per block.

In this paper we present new results relative to the "expectation-maximization/maximization of the posterior marginals" (EM/MPM) algorithm for simultaneous parameter estimation and segmentation of textured images. The EM/MPM algorithm uses a Markov random field model for the pixel class labels and alternately approximates the MPM estimate of the pixel class labels and estimates parameters of the observed image model. The goal of the EM/MPM algorithm is to minimize the expected value of the number of misclassified pixels. We present new theoretical results in this paper which show that the algorithm can be expected to achieve this goal, to the extent that the EM estimates of the model parameters are close to the true values of the model parameters. We also present new experimental results demonstrating the performance of the EM/MPM algorithm.

We present a new algorithm for segmentation of textured images using a multiresolution Bayesian approach. The new algorithm uses a multiresolution Gaussian autoregressive (MGAR) model for the pyramid representation of the observed image, and assumes a multiscale Markov random field model for the class label pyramid. The models used in this paper incorporate correlations between different levels of both the observed image pyramid and the class label pyramid. The criterion used for segmentation is the minimization of the expected value of the number of misclassified nodes in the multiresolution lattice. The estimate which satisfies this criterion is referred to as the "multiresolution maximization of the posterior marginals" (MMPM) estimate, and is a natural extension of the single-resolution "maximization of the posterior marginals" (MPM) estimate. Previous multiresolution segmentation techniques have been based on the maximum a posterior (MAP) estimation criterion, which has been shown to be less appropriate for segmentation than the MPM criterion. It is assumed that the number of distinct textures in the observed image is known. The parameters of the MGAR model-the means, prediction coefficients, and prediction error variances of the different textures-are unknown. A modified version of the expectation-maximization (EM) algorithm is used to estimate these parameters. The parameters of the Gibbs distribution for the label pyramid are assumed to be known. Experimental results demonstrating the performance of the algorithm are presented.

In this paper we present new results relative to the "expectation-maximization/maximization of the posterior marginals" (EM/MPM) algorithm for simultaneous parameter estimation and segmentation of textured images. The goal of the EM/MPM algorithm is to minimize the expected value of the number of misclassified pixels. We present new theoretical results in this paper which show that the algorithm can be expected to achieve this goal, to the extent that the EM estimates of the model parameters are close to the true values of the model parameters. We also present new experimental results demonstrating the performance of the algorithm.

In this paper we present a new algorithm for segmentation of noisy or textured images using a multiresolution Bayesian approach. Our algorithm is different from previously proposed multiresolution segmentation techniques in that we use a multiresolution Gaussian autoregressive (AR) model for the pyramid representation of the observed image. Our algorithm also approximates the "maximization of the posterior marginals" (MPM) estimate of the pixel class labels at each resolution, from coarsest to finest, unlike previously proposed techniques, which have been based on MAP estimation. Experimental results are presented to demonstrate the performance of the new algorithm.

Presents a new algorithm that utilizes mathematical morphology for pyramidal coding of color images. The authors obtain lossy color image compression by using block truncation coding at the pyramid levels to attain reduced data rates. The pyramid approach is attractive due to low computational complexity, simple parallel implementation, and the ability to produce acceptable color images at moderate data rates. In many applications, the progressive transmission capability of the algorithm is very useful. The authors show experimental results for color images at data rates of 1.89 bits/pixel.

Presents a new algorithm for segmentation of noisy or textured images using the expectation-maximization (EM) algorithm for estimating parameters of the probability mass function of the pixel class labels and the maximization of the posterior marginals (MPM) criterion for the segmentation operation. A Markov random field (MRF) model is used for the pixel class labels. The authors present experimental results demonstrating the use of the new algorithm on synthetic images and medical imagery.>

Multiresolution image decompositions (e.g., wavelets), in conjunction with a variety of quantization schemes, have been shown to be very effective for image compression. Recently, several promising tree-structured quantization schemes that exploit the correlation across scales have been proposed. In this paper, we present an image compression algorithm based on a multiresolution Markov random field model used to model the correlations of wavelet coefficients across the scales. We also present experimental results obtained using the algorithm.>