Modern data-centric workloads are driving rapid growth in the demand for high-bandwidth, large-capacity memory. Compute Express Link (CXL) has emerged as a key technology for scalable memory expansion through the Peripheral Component Interconnect Express (PCIe)'s high-speed serial interfaces. However, while PCIe 6.0's L0p mode enables lane-level power gating without disrupting traffic flow, it lacks a policy mechanism for dynamically deciding the number of active lanes. Since actual energy savings depend on matching active lanes to runtime bandwidth demand, such a policy is essential. This paper presents DDLM (Demand-Aware Dynamic Link Width Management), a lightweight control framework that improves CXL link energy efficiency by dynamically adjusting link width based on runtime traffic. DDLM integrates two complementary modules: a predictor that predicts bandwidth demand based on short-term temporal locality to set the width ahead of hardware transition latency, and a congestion monitor that detects transient congestion via internal queue inspection. These modules are coordinated by a finite state machine that manages safe link width transitions under physical-layer constraints. We implement DDLM in the Ramulator2 simulator and evaluate it with diverse SPEC CPU workloads. Compared to a fixed-width baseline, DDLM reduces CXL Memory energy by up to 13%, improves utilization by$2.22 \times$, and limits performance loss to under 3%. This work offers a practical path to energy-proportional CXL Memory.

Building an ideal graph which reveals the exact intrinsic structure of the data is critical in graph-based clustering. There have been a lot of efforts to construct an affinity matrix satisfying such a need in terms of a similarity measure. A recent approach attracting attention is on using doubly stochastic normalization of the affinity matrix to improve the clustering performance. In this paper, we propose a novel method to build a high-quality affinity matrix via incorporating Davis-Kahan theorem of matrix perturbation theory in the doubly stochastic normalization problem. We interpret the goal of the doubly stochastic normalization problem as minimizing the relative distance between the eigenspaces of the corresponding matrices. Also, for the doubly stochastic normalization problem we include an additional constraint that each eigenvalue be on the unit interval to fully conform to the spectral graph theory. Experiments on our framework present superior performance over various datasets.

In this paper, a hierarchical motion estimation (ME) algorithm is proposed for motion-compensated frame interpolation. The algorithm estimates the true motion vector field (MVF) of a video frame from its candidate MVFs, which are the results of full-search block-matching that utilizes multiple block sizes, by maximizing the posterior probability for the true MVF. Owing to probabilistic models utilized for defining the posterior probability, the estimate of the true MV of a pixel block is not only largely affected by the most reliable MV among the candidate MVs but also effectively corrected by its spatially neighboring MVs when all the candidate MVs turn out to be unreliable. Experimental results showed that the proposed algorithm outperforms existing hierarchical ME algorithms in terms of the quality of interpolated frames.

In three-dimensional display systems, depth image based rendering is the most commonly used technique for generating images captured from a virtual viewpoint through reference views and depth maps. However, disoccluded hole filling remain a challenging issue as newly exposed area appears in the virtual view. Image inpainting based hole filling is a popular approach for filling in the hole, but the conventional methods generate annoying artifacts along the boundaries of the foreground objects. This letter presents a robust hole filling method that generates a virtual view with high quality. First, by using the local and neighboring property of the depth map, the background region is filled prior to the foreground region. Also, in the matching step to find the data to fill in, depth homogeneity is considered. Finally, based on the boundary decision using the depth property, the boundary regions are filled with background-relevant data to reduce boundary artifacts. Experimental results show that the views synthesized by the proposed method achieve higher visual comfort than the view synthesized by the existing methods.

A new motion vector field (MVF) refinement algorithm is proposed for block-based motion-compensated frame interpolation. Under the assumption that an observed MVF, such as the result of a block-based motion estimation (BME), is a degraded version of the true MVF, the true MVF is estimated from the observation through maximum a posteriori probability (MAP) estimation. To define the posterior probability of the true MVF, the degradation is modeled as a locally stationary additive Gaussian noise, so the variance of the noise represents the unreliability of the observed motion vector (MV). The noise variance is directly estimated from the observation vector and its select neighbors. The prior distribution of the true MVF is designed to rely on the distances between the MV and its neighbors and to properly smooth the false MVs in the observation. The MAP estimate of the true MVF is obtained via the iterative conditional mode method. The outcome is a set of iterative update equations, which produce the kth estimate of the true MV of a block by combining, according to the estimated noise variance, the observation and the neighboring (k - 1)th estimates. Experimental results prove that the proposed algorithm achieves performances comparable with those of several existing MAP-based BME algorithms at a much lower computational complexity.

This paper proposes a new approach to motion vector field smoothing for block-based motion-compensated frame interpolation (MCFI). Based on the assumption that an observed motion vector field, which is the result of a block-based motion estimation (BME), is a degraded version of the true motion vector field, we calculate the maximum a posteriori (MAP) estimate of the true motion vector field from the observed. The degradation and the true motion vector field are modeled as additive Gaussian noise and a Markov random field, respectively. Iterative conditional modes (ICM) method is used for calculating the MAP estimate. The experimental results show that the proposed algorithm not only smoothes MVFs but also preserves motion boundaries better than the existing methods.

In order to reduce the blocking artifact in the Joint Photographic Experts Group (JPEG)-compressed images, a new noniterative postprocessing algorithm is proposed. The algorithm consists of a two-step operation: low-pass filtering and then predicting. Predicting the original image from the low-pass filtered image is performed by using the predictors, which are constructed based on a broken line regression model. The constructed predictor is a generalized version of the projector onto the quantization constraint set, or the narrow quantization constraint set. We employed different predictors depending on the frequency components in the discrete cosine transform (DCT) domain since each component has different statistical properties. Further, by using a simple classifier, we adaptively applied the predictors depending on the local variance of the DCT block. This adaptation enables an appropriate blurring depending on the smooth or detail region, and shows improved performance in terms of the average distortion and the perceptual view. For the major-edge DCT blocks, which usually suffer from the ringing artifact, the quality of fit to the regression model is usually not good. By making a modification of the regression model for such DCT blocks, we can also obtain a good perceptual view. The proposed algorithm does not employ any sophisticated edge-oriented classifiers and nonlinear filters. Compared to the previously proposed algorithms, the proposed algorithm provides comparable or better results with less computational complexity.

This letter proposes a new asymmetric watermarking system. In the proposed system, a linear transformation A is applied to the primitive key u to form the private encoding key Au and the public decoding key A/sup -t/u, where -t denotes inverse transpose. The transformation mostly cancels out when the decoding key is correlated to watermarked signal containing the encoding key, which is possibly contaminated with noise. Our analysis shows that this simple scheme provides not only reliability in watermark detection but also security from any attempts to identify or erase the "encoding" key. We also address the issue of selecting a good transformation matrix. The proposed system can be made fit for each application by properly designing the transform.

In this letter, we propose a new asymmetric watermarking system which can accommodate many embedding watermarks but needs only one reference watermark for detection. Such a system is useful in averting attacks that seek to estimate the embedding watermark. In the proposed system, the phase of the reference watermark is shifted randomly (clockwise or counterclockwise) in the discrete Fourier transform domain to make embedding watermarks. They are correlated with one another and have the same correlation with the reference one. We also address how to select the design parameters.

This paper presents a new double-talk detector (DTD) based on echo path estimation. The proposed algorithm consists of two stages. In the £rst stage, single-talk periods are distinguished from the double-talks or echo path changes based on the energy level of the echo path estimate. An accurate distinction between the double-talk and echo path change is made in the second stage based on the gradient of the energy level in the estimated echo path. By experiments, it is found that the proposed approach is effective in detecting double-talk periods. Moreover, the required decision delay is shorter than that of the conventional methods.

In this paper, we propose a new postprocessing algorithm that reduces the blocking artifacts in low-rate coded images. It consists of two-step operations: low-pass £ltering and image estimation. The latter makes an estimation of the original image from the £ltered image based on regression functions. Regression functions for the JPEG-coded real images are numerically evaluated from a training set, and their piecewise linear approximation are used for the estimation. This approximate unbiased estimator is applied adaptively, depending on the DCT coef£cients. This proposed approach is a generalization of the existing methods as QCS [6] and NQCS [3], [4] can be regarded as the same type that employ biased estimators. Simulation results show that the new algorithm outperforms the existing methods in both objective and subjective qualities.

We propose a post-processing method in the wavelet transform domain that can significantly reduce the blocking effects in low-bit-rate block-transform-coded images. Although the quantization noise of the transform coefficients is the sole source of error in a coded image, the properties of block transform make the errors appear in two categories: blocky noise, which causes blocking effects, and granular (nonblocky) noise. Noting that subband coding does not suffer from blocky noise, the proposed technique is designed to work in the subband domain. Once a coded image is decomposed into subbands by wavelet filters, most energy of the blocky noise exists on the predetermined block boundaries of their corresponding subbands. We can reduce the blocky noise by a linear minimum mean square error filter, which fully exploits the characteristics of the signal and noise components in each subband. After the blocky noise is reduced, the granular noise can further be decreased by exploiting its nonstructuredness. Computer simulations show that the proposed method visibly reduces the blocking effects in reconstructed images and yields better PSNR improvement.

This paper proposes a postprocessing algorithm that can reduce the blocking artifacts in discrete cosine transform (DCT) coded images. To analyze blocking artifacts as noise components residing across two neighboring blocks, we use 1-D pixel vectors made of pixel rows or columns across two neighboring blocks. We model the blocky noise in each pixel vector as a shape vector weighted by the boundary discontinuity. The boundary discontinuity of each vector is estimated from the difference between the pixel gradient across the block boundary and that of the internal pixels. We make minimum mean squared error (MMSE) estimates of the shape vectors, indexed by the local image activity, based on the noise statistics prior to postprocessing. Once the estimated shape vectors are stored in the decoder, the proposed algorithm eliminates the noise components by simply subtracting from each pixel vector an appropriate shape vector multiplied by the boundary discontinuity. The experimental results show that the proposed algorithm is highly effective in reducing blocking artifacts in both subjective and objective viewpoints, at low computational burden.

This paper discusses a criterion for testing a vector quantizer (VQ) codebook that is obtained by "training". When a VQ codebook is designed by a clustering algorithm using a training set, "time-average" distortion, which is called the training-set-distortion (TSD), is usually calculated in each iteration of the algorithm, since the input probability function is unknown in general and cumbersome to deal with. The algorithm stops when the TSD ceases to significantly decrease. In order to test the resultant codebook, validating-set-distortion (VSD) is calculated on a separate validating set (VS). Codebooks that yield small difference between the TSD and the VSD are regarded as good ones. However, the difference VSD-TSD is not necessarily a desirable criterion for testing a trained codebook unless certain conditions are satisfied. A condition that is previously assumed to be important is that the VS has to be quite large to well approximate the source distribution. This condition implies greater computational burden of testing a codebook. In this paper, we first discuss the condition under which the difference VSD-TSD is a meaningful codebook testing criterion. Then, convergence properties of the VSD, a time-average quantity, are investigated. Finally we show that for large codebooks, a VS size as small as the size of the codebook is sufficient to evaluate the VSD. This paper consequently presents a simple method to test trained codebooks for VQ's. Experimental results on synthetic data and real images supporting the analysis are also provided and discussed.

This correspondence formally defines partial Radon trans- forms for functions of more than two dimensions. It shows that a generalized projection-slice theorem exists which connects planar and hy- perplanar projections of a function to its Fourier transform. In addition, a general theoretical framework is provided for carrying out -dimensional backprojection reconstruction in a multistage fashion through the use of the partial Radon transform.

A class of vector quantizers with memory that are known as finite state vector quantizers (FSVQs) in the image coding framework is investigated. Two FSVQ designs, namely side match vector quantizers (SMVQs) and overlap match vector quantizers (OMVQs), are introduced. These designs take advantage of the 2-D spatial contiguity of pixel vectors as well as the high spatial correlation of pixels in typical gray-level images. SMVQ and OMVQ try to minimize the granular noise that causes visible pixel block boundaries in ordinary VQ. For 512 by 512 gray-level images, SMVQ and OMVQ can achieve communication quality reproduction at an average of 1/2 b/pixel per image frame, and acceptable quality reproduction. Because block boundaries are less visible, the perceived improvement in quality over ordinary VQ is even greater. Owing to the structure of SMVQ and OMVQ, simple variable length noiseless codes can achieve as much as 60% bit rate reduction over fixed-length noiseless codes.

Two finite-state vector designs are introduced, namely, side-match vector quantizers (SMVQ) and overlap-match vector quantizers (OMVQ). The underlying ideas of SMVQ and OMVQ are simple and intuitive, and so are their design procedures. They try to avoid large grey-level transitions across pixel block (vector) boundaries which cause the annoying 'blockyness' effect in memoryless VQ. Experimental results show that when they are applied to 512*512 grey-level images, SMVQ and OMVQ can achieve communication-quality reproduction 33.5 dB peak-signal-to-noise ratio (PSNR) at the average rate of 1/2 bit per pixel per image frame and acceptable quality reproduction (30 dB PSNR) at 1/4 bit per pixel. These performances are superior to those of ordinary VQ by more than 3 dB in PSNR. Further, because block boundaries are less visible, the perceived improvement in quality over ordinary VQ is even greater.>

A new class of chain codes for line drawings, named delta codes is introduced. Two specific types of fixed angle delta codes and ring delta codes are described. The area-of-error distortion and rate are calculated for two types of delta codes applied to straight and slowly curving lines, and it is shown that the delta outperforms other chain codes with the same set of link types. For rapidly curving lines, however, the delta codes introduce curvature overload distortion analogous to the slope overload distortion of traditional delta modulation for waveforms.>