Search

We present an improved directional weighted interpolation method for single-sensor camera imaging. By observing the fact that the conventional directional weighted interpolation methods are based on unreliable assumptions using spectral correlation, a contribution of this work is made using an antialiasing finite impulse response filter to improve the interpolation accuracy by exploiting robust spectral correlation. We also make a contribution toward refining the interpolation result by using the gradient inverse weighted filtering method. An experimental analysis of images revealed that our proposed algorithm provided superior performance in terms of both objective and subjective image quality compared to conventional directional weighted demosaicking algorithms. Our implementation has very low complexity and is, therefore, well suited for real-time applications.

A tone reproduction algorithm with enhanced contrast of high-dynamic range images on conventional low-dynamic range display devices is presented. The proposed algorithm consists mainly of block-based parameter estimation, a characteristic-based luminance adjustment, and an adaptive Gaussian filter using minimum description length. Instead of relying only on the reduction of the dynamic range, a characteristic-based luminance adjustment process modifies the luminance values. The Gaussian-filtered luminance value is obtained from appropriate value of variance, and the contrast is then enhanced through the use of a relation between the adjusted luminance and Gaussian-filtered luminance values. In the final tone-reproduction process, the proposed algorithm combines color and luminance components in order to preserve the color consistency. The experimental results demonstrate that the proposed algorithm achieves a good subjective quality while enhancing the contrast of the image details.

We propose a local region statistics-based weighted interpolation filter for intrafield deinterlacing. The proposed algorithm consists of three steps: first, we preinterpolate the missing line with an efficient interpolation filter in the working window. Then, we calculate the coefficients of the center-weighted interpolation filter by exploiting local statistics, namely the center-weighted mean, the center-weighted variance, and the closeness of the neighboring pixels. In the last step, we interpolate the missing line using the proposed filter. Experimental results show that the proposed algorithm provides superior performances in terms of both objective and subjective image qualities than the existing algorithm.

A satellite image enhancement method based on singular value decomposition (SVD) is presented. The proposed method consists mainly of four steps: image decomposition, adjustable contrast enhancement, noise reduction, and image composition. The method decomposes an image into its singular values, from which the luminance information can be obtained using SVD. In turn, adjustable contrast enhancement is applied by changing the singular values, which are controlled by characteristic-based contrast-level parameters. In the noise reduction process, the high-frequency elements, such as noise, are removed using rank approximation. In the final image composition process, the proposed algorithm combines color and luminance components in order to preserve color consistency. Experimental results show that the proposed scheme has less noise than the conventional methods and improves contrast performance while preserving details.

An advanced two-step motion estimation algorithm that uses adjustable partial distortion for fast motion estimation is proposed. To over- come the demerits of existing two-step search algorithms, we propose a new, fast two-step algorithm, which achieves both speed and peak signal- to-noise ratio (PSNR) improvement by adding two predicted points. We improve the two-step search by using the relationship between a neighbor block and the current block. The proposed algorithm is 187 times faster than full search and 2.7 times faster than two-step edge based A partial distance search with negligible PSNR degradation. Therefore it is suitable for real-time video implementation. © 2013 Society of Photo-Optical Instrumentation

Our purposeis to present an intrafield deinterlacing method using the Bayes classifier. The conventional intrafield deinterlacing methods interpolate the pixel along the local edge direction, but they yield interpolation errors when the local edge direction is deter- mined to be wrong. On the basis of the Bayes classifier, the proposed algorithm performs region-based deinterlacing. The proposed algo- rithm utilizes an input feature vector that includes five directional cor- relations, which are used to extract the characteristics of the local region, to classify the local region. After the classification of the local region, one of the three simple interpolation methods, which pos- sesses the highest probability to be used among the three, is chosen for the corresponding local region. In addition, we categorized the range of the feature vector to reduce the computational complexity. Simulation results show that the proposed Bayes classifier-based deinterlacing method minimizes interpolation errors. Compared to the traditional deinterlacing methods and Wiener filter-based interpo- lation method, the proposed method improves the subjective quality of the reconstructed image, and maintains a higher peak signal-to-noise ratio level. © 2013 SPIE and IS&T (DOI: 10.1117/1.JEI.22.2.023003)

In H.264/advanced video coding (AVC), low-pass filtering is performed on neighboring coded pixels before intra prediction is applied to reduce the effect of quantization errors. However, reducing the effect of quantization errors by conventional methods is difficult in some areas, depending on the image characteristics. To alleviate this issue, we need to consider the prediction direction. We propose a mode-dependent filtering technique to reduce the effect of quantization errors. Our method has a threshold to adaptively choose between the conventional and proposed methods to reduce the number of flag bits that are sent to the decoder. The method is compared with other benchmark methods in terms of their average bit-saving and peak signal-to-noise ratio. Simulation results show that the average bit-rate reduction is about 1.222% in test sequences.

This paper proposes an efficient intra-field deinterlacing algorithm which implements a filter switching interpolation method (FSID) according to different local region features, which produces outstanding visual results. In this method, a bilateral filter is used for smooth regions to interpolate the missing line due to its accurate interpolation performance through use of the closeness and similarity among the neighboring pixels. On the other hand, a trilateral filter is used for complex regions to interpolate the missing line since it has high edge-preserving ability due to taking into accountthe closeness and similarity among neighboring pixels, but also the pixel gradient. Experimental results show that the proposed FSID algorithm provides satisfactory performances in terms of both objective and subjective image quality. In addition, the algorithm exploits the local spatial similarity among the neighboring pixels without complex preset-conditions, which is easier to implement than many other existing algorithms.

A new stereoscopic satellite image enhancement method based on singular value decomposition (SVD) is presented. The proposed stereoscopic image enhancement technique consists of four steps; SVD based on image analysis, image enhancement, noise reduction, and tone consistency. The goal of this work is to enhance low-contrast stereo satellite images with noise elimination, and obtain identical luminance levels from slightly different luminance in stereo images. Experimental results show the proposed method improves contrast performance while maintaining the luminance continuity. It also exhibits less noise than conventional methods.

We propose a discrete cosine transform (DCT) analysis-based adaptive deblocking algorithm to reduce the block discontinuities. First we analyze the properties of the one-dimensional DCT coefficients. Based on this analysis, we utilize two adjacent vectors to classify the vector-containing block boundary into three models. For each model, we apply different filters according to the local properties. Comprehensive experiments indicate that the proposed method outperforms a large number of deblocking methods in the literature in terms of the evaluation of peak signal-to-noise ratio, perceptual quality, and computational complexity.

We propose an adaptive search range adjustment algorithm that is particularly suitable for low-bit-depth motion estimation approaches. Typical low-bit-depth motion estimation approaches, such as one-bit transform (1BT), two-bit transform (2BT), and truncated gray-coded bit-plane matching (TGCBM), perform a full search with low-bit expression. We propose combining a search-range adjustment, which can be predicted by using motion vector information of the reference frame and the existing bit-wise matching methods, such as 1BT, 2BT, and TGCBM. Combining this proposed algorithm and low-bit expression enhances both motion estimation accuracy and time consumption. Experimental results show that low-bit-depth motion estimation approaches using this proposed algorithm have an outstanding performance compared with those without the proposed algorithm both in motion estimation accuracy and time consumption.

A widely used contrast enhancement method, the histogram equalization (HE) often produces images with unnatural appearances and visually disturbing artifacts because the HE compels the enhanced image to follow the uniform distribution. An adaptive histogram equalization using logarithmic mapping is presented, with a proposed algorithm based on a bin underflow and overflow method that achieves contrast enhancement by putting constraints on each histogram component differently. To incorporate characteristics of the human visual system, the logarithmic mapping function is used as constraint function, while the rate of contrast enhancement is controlled by determining the control parameters with the characteristics of the original image. The experimental results show that the proposed algorithm not only keeps the original histogram shape features, but also enhances the contrast effectively. Due to its simplicity, the proposed algorithm can be applied by simple hardware and processed in a real-time system.

We present an enhanced motion estimation and compensation algorithm by prefiltering reference frames before motion estimation. The conventional block based motion estimation algorithm gives poor performance when abrupt motion change occurs. The proposed algorithm constructs prefilters based on motion vector distribution analysis and compensates temporal sampling artifacts, such as blur or deblur, between adjacent frames. Compared to H.264/AVC, the proposed algorithm achieves significant bit-rate reduction up to 14.59%.

We introduces an efficient intra-field deinterlacing algorithm while considering both closeness and similarity between the interpolated pixel and the neighbor pixels. Instead of estimating the edge orientations via limited candidate directions as previous intra-field deinterlacing methods, we propose an adaptive spatial locality- and similarity-based deinterlacing method. Many deinterlacing methods described in the literature only consider the similarity between neighbor pixels and interpolated pixels; however, awareness of distance correlation and spatial locality has not yet been studied. Experimental results indicate that the proposed scheme outperforms a number of existing approaches, in terms of both the objective and subjective performance.

This paper proposes an efficient deinterlacing algorithm using the approximated sinc interpolator and adaptive distance weighting scheme that is suitable for real-time applications. The conventional edge-based deinterlacing algorithms provide poor image quality due to the limitation of the number of edge directions considered; moreover, the existing algorithms which consider many edge directions have high computational burden while still producing the staircase artifacts at the complex region. To overcome these problems, the proposed algorithm adopts an approximated sinc interpolator and adaptive distance weighting scheme which utilizes the pixel correlation. Simulation results show that the proposed algorithm provides outstanding performances in terms of both the objective and subjective image qualities, and it has lower computational complexity than the existing algorithms.

In this paper, we introduce a direction-oriented covariance-based adaptive deinterlacing (CAD) method. The conventional CAD method preserves edge direction; however, it does not consider the local edge direction, which makes it unsuitable for real-time applications. The proposed method only utilizes the neighbor pixels aligned in the local edge direction to obtain optimal coefficients, with less processing time than the aforementioned CAD. First, the local edge direction is determined using the modified edge-based line average method. Then, based on direction-oriented geometric duality, the optimal interpolation coefficients for the neighbor pixels in the corresponding direction are estimated using the Wiener filter. Furthermore, we present inverse-free least squares approximations using orthogonal-triangular decomposition based on the Gram-Schmidt process. This technique's performance improves as the size of the image becomes larger. The experimental results prove that the proposed method provides a significant improvement over other deinterlacing methods.

This paper proposes a single-field deinterlacing method based on the autoregressive model and edge map. The new method interpolates missing pixels through estimating the deinterlaced covariance from the interlaced covariance, instead of estimating the edge orientations as previous intrafield deinterlaced methods (line average, edge-based line-average, direction-oriented interpolation, etc.) do. The proposed method adopts autoregressive mechanism, which considers mutual influence between the estimated missing pixels in a slip window. In addition, adding an edge map in our algorithm is used to reduce the computational complexity. The experimental results show that the proposed method outperformed the previous method in peak signal-to-noise ratio, and common artifacts (serration, line crawl, flicker, blurring, etc.) are significantly reduced.

We present an efficient deinterlacing method via mathematical modeling of the neighbor pixels in the local region. The local surface model is designed using the quadratic equation having two-dimensional coordinate variables. Unlike conventional deinterlacing methods, the proposed method avoids using directional difference measures, resulting in reduced limitation on the number of considering edge directions. By modeling the local surface, it is easier to derive the true characteristic of the local region in a natural image than utilizing the directional difference measure. In order to decide the optimal coefficients of the surface model, the neighbor pixels around the current pixel to be interpolated are utilized. Once the coefficients are determined, the surface model estimates the pixel intensity of the current pixel to be interpolated. Simulation results show that the proposed surface model-based deinterlacing method minimizes the interpolation error. Compared to the traditional deinterlacing methods and Wiener filter-based interpolation method, the proposed method improves the subjective quality of the interpolated edges while maintaining a higher peak signal-to-noise-ratio level.

This paper proposes a refinement algorithm for equivalent numbers of nonmatching points on surfaces for binary-block motion estimation. The conventional binary-block motion estimation algorithms use the number of nonmatching points as a distortion criterion, and that limits the dynamic range of the distortion. The proposed algorithm performs refinement using the sum of absolute differences on surfaces that have equal numbers of nonmatching points. The refinement step improves the visual quality of binary-block motion estimation at the cost of higher computational complexity.

In this paper, we propose an embedded satellite image compression method using Weighted ZeroBlock Coding (WZBC) and optimal sorting. In order to reduce average codeword length, Set Partition Embedded block (SPECK) and Embedded ZeroBlock Coder (EZBC) both encode significant block-sets with fixed-length bits, while WZBC assigns different-length bits to encode block-sets which contain different numbers of significant subblocks. In view of the context correlation among coefficients/blocks, WZBC employs a weight context to optimize the scanning order of the significance testing and the ratedistortion performance. Experimental results show that the proposed WZBC in binary coding mode provides excellent coding performance compared with those of SPECK and Set Partitioning In Hierarchical Trees (SPIHT) which use arithmetic coding, and can even closely approach that of JPEG2000. When arithmetic coding is extensively used, the proposed method has clear advantages.

We propose a robust and fast quarter-pixel motion estimation algorithm. This algorithm is an advanced version of the previously proposed model-based quarter-pixel motion estimation (MBQME). MBQME has many advantages in computational complexity, running speed, and hardware implementations. But it has the problem that it does not find the quarter-pixel positions that locate beyond the half-pixel positions. That is one of limitations of model-based motion estimation methods, and it leads to both peak-SNR degradation and bit-rate increase. To solve this problem, we propose a hierarchical mathematical model with minimum interpolations. Through this model, we can determine a motion vector at every quarter-pixel point, which is perfectly compatible with the quarter-pixel motion estimation method within international video coding standards such as MPEG-4 and H.264/AVC. The simulation results show that the proposed method yields almost the same or even better peak-SNR performance than that of full-search quarter-pixel motion estimation, with much lower computational complexity.

We discuss deinterlacing results in a computationally constrained and varied environment. The proposed computation-aware algorithm selection approach (CASA) for fast interlaced to progressive conversion algorithm consists of three methods: the line-averaging (LA) method for plain regions, the modified edge-based line-averaging (MELA) method for medium regions, and the proposed covariance-based adaptive deinterlacing (CAD) method for complex regions. The proposed CASA uses two criteria, mean-squared error (MSE) and CPU time, for assigning the method. We proposed a CAD method. The principle idea of CAD is based on the correspondence between the high and low-resolution covariances. We estimated the local covariance coefficients from an interlaced image using Wiener filtering theory and then used these optimal minimum MSE interpolation coefficients to obtain a deinterlaced image. The CAD method, though more robust than most known methods, was not found to be very fast compared to the others. To alleviate this issue, we proposed an adaptive selection approach using a fast deinterlacing algorithm rather than using only one CAD algorithm. The proposed hybrid approach of switching between the conventional schemes (LA and MELA) and our CAD was proposed to reduce the overall computational load. A reliable condition to be used for switching the schemes was presented after a wide set of initial training processes. The results of computer simulations showed that the proposed methods outperformed a number of methods presented in the literature.

The H.264/AVC (advanced video coding) is used in a wide variety of applications including digital broadcasting and mobile applications, because of its high compression efficiency. The variable block mode scheme in H.264/AVC contributes much to its high compression efficiency but causes a selection problem. In general, rate-distortion optimization (RDO) is the optimal mode selection strategy, but it is computationally intensive. For this reason, the H.264/AVC encoder requires a fast mode selection algorithm for use in applications that require low-power and real-time processing. A probable mode prediction algorithm for the H.264/AVC encoder is proposed. To reduce the computational complexity of RDO, the proposed method selects probable modes among all allowed block modes using removable SKIP mode distortion estimation. Removable SKIP mode distortion is used to estimate whether or not a further divided block mode is appropriate for a macroblock. It is calculated using a no-motion reference block with a few computations. Then the proposed method reduces complexity by performing the RDO process only for probable modes. Experimental results show that the proposed algorithm can reduce encoding time by an average of 55.22% without significant visual quality degradation and increased bit rate.

We present an adaptive algorithm that finds the best block-matching results in a computationally constrained and varied environment. The conventional diamond search algorithm, though faster than most known algorithms, is not very robust for sequences with scene variations or significant global motion. To solve this issue, rather than only using one fast motion estimation algorithm, we devise a more adaptive selection of fast motion estimation algorithms. Our adaptive selection approach for fast block search (ASFBS) algorithm uses a diamond search and two new subalgorithms: a cross-three-step search algorithm for large moving images and an advanced cross-diamond search algorithm for small moving images. The proposed ASFBS adapts based on the length of the motion vector, the number of search points, and the matching criteria of the neighboring block. Experimental results show that ASFBS is much more robust; it is faster than other popular fast block-matching algorithms, with smaller distortions.

We present an adaptive tone reproduction algorithm for displaying high-dynamic-range (HDR) images on conventional low-dynamic-range (LDR) display devices. The proposed algorithm consists of an adaptive tone reproduction operator and chromatic adaptation. The algorithm for dynamic range reduction relies on suitable tone reproduction functions that depend on histogram-based parameter estimation to adjust luminance according to global and local features. Instead of relying only on reduction of dynamic range, this chromatic adaption technique also preserves chromatic appearance and color consistency across scene and display environments. Our experimental results demonstrate that the proposed algorithm achieves good subjective quality while preserving image details. Furthermore, the proposed algorithm is simple and practical for implementation.

The H.264/AVC baseline profile is used in many applications, including digital multimedia broadcasting, Internet protocol television, and storage devices, while the MPEG-2 main profile is widely used in applications, such as high-definition television and digital versatile disks. The MPEG-2 main profile supports B pictures for bidirectional motion prediction. Therefore, transcoding the MPEG-2 main profile to the H.264/AVC baseline is necessary for universal multimedia access. In the cascaded pixel domain transcoder architecture, the calculation of the rate distortion cost as part of the mode decision process in the H.264/AVC encoder requires extremely complex computations. To reduce the complexity inherent in the implementation of a real-time transcoder, we propose a fast mode decision algorithm based on complexity information from the reference region that is used for motion compensation. In this study, an adaptive mode decision process was used based on the modes assigned to the reference regions. Simulation results indicated that a significant reduction in complexity was achieved without significant degradation of video quality.

Variable-length coding (VLC) is widely used in video coding to improve compression efficiency. However, suffering from loss of synchronization, the VLC bit stream is much more sensitive to random errors than a fixed-length coding (FLC) bit stream. Error-resilient entropy coding (EREC) is a valid tool combating random errors in a VLC bit stream. Due to its intrinsic property of error propagation, when EREC is applied to a video bit stream, those blocks placed later become much more likely to be lost. We propose a simple method to further improve the error robustness of a video bit stream by interleaving transform coefficients of blocks so that low-frequency information is always placed ahead of high-frequency information. Thus, low-frequency information of greater significance is less likely to be lost. Experimental results prove the superiority of the proposed method. In addition, block interleaving can also be used in a data-partitioned video bit stream with ease.