Search

This study details the development of a lightweight and high performance model, targeting real-time object detection. Several designed features were integrated into the proposed framework to accomplish a light weight, rapid execution, and optimal performance in object detection. Foremost, a sparse and lightweight structure was chosen as the network’s backbone, and feature fusion was performed using modified feature pyramid networks. Recent learning strategies in data augmentation, mixed precision training, and network sparsity were incorporated to substantially enhance the generalization for the lightweight model and boost the detection accuracy. Moreover, knowledge distillation was applied to tackle dropping issues, and a student–teacher learning mechanism was also integrated to ensure the best performance. The model was comprehensively tested using the MS-COCO 2017 dataset, and the experimental results clearly demonstrated that the proposed model could obtain a high detection performance in comparison to state-of-the-art methods, and required minimal computational resources, making it feasible for many real-time deployments.

Mitochondria are the organelles that generate energy for the cells. Many studies have suggested that mitochondrial dysfunction or impairment may be related to cancer and other neurodegenerative disorders such as Alzheimer’s and Parkinson’s diseases. Therefore, morphologically detailed alterations in mitochondria and 3D reconstruction of mitochondria are highly demanded research problems in the performance of clinical diagnosis. Nevertheless, manual mitochondria segmentation over 3D electron microscopy volumes is not a trivial task. This study proposes a two-stage cascaded CNN architecture to achieve automated 3D mitochondria segmentation, combining the merits of top-down and bottom-up approaches. For top-down approaches, the segmentation is conducted on objects’ localization so that the delineations of objects’ contours can be more precise. However, the combinations of 2D segmentation from the top-down approaches are inadequate to perform proper 3D segmentation without the information on connectivity among frames. On the other hand, the bottom-up approach finds coherent groups of pixels and takes the information of 3D connectivity into account in segmentation to avoid the drawbacks of the 2D top-down approach. However, many small areas that share similar pixel properties with mitochondria become false positives due to insufficient information on objects’ localization. In the proposed method, the detection of mitochondria is carried out with multi-slice fusion in the first stage, forming the segmentation cues. Subsequently, the second stage is to perform 3D CNN segmentation that learns the pixel properties and the information of 3D connectivity under the supervision of cues from the detection stage. Experimental results show that the proposed structure alleviates the problems in both the top-down and bottom-up approaches, which significantly accomplishes better performance in segmentation and expedites clinical analysis.

Digital halftoning deals with transforming a gray or color image into its binary version which is useful in printing applications. Dot diffusion is one of the prominent halftone methods which can yield superior image quality with parallel processing capabilities. In this paper, a rapid watermarking algorithm is proposed for dot-diffusion halftone images using adaptive class-matrix selection and modified error diffusion kernels. To process the image using an adaptive class matrix, the processing order of the class matrix is reversed and transposed, and for error diffusion the coefficients are replaced with different weights. For decoding, an effective strategy is proposed based on a correlation analysis and halftone statistics. The proposed strategy can successfully embed and decode the binary watermark from a single dot-diffused halftone image. From the experimental results, the proposed method is found to be effective in terms of good decoding accuracy, imperceptibility and robustness against various printed distortions.

This paper presents a new method for denoising an image corrupted with Additive White Gaussian Noise (AWGN). We adopt the effectiveness of Convolutional Vision Transformer (CvT) to suppress the occurred noise on an image. The proposed method exploits the residual learning approach in order to estimate and reduce the noise on a noisy image. Herein, the model is trained in the end-to-end manner to capture the relation between the noisy image and its noise map. The experimental results show the effectiveness of the proposed method in terms of subjective and objective measurements.

In this paper, a new night dehazing strategy is proposed by modifying the transmission and atmospheric light conditions. To suppress the local light sources, a Gaussian based atmospheric light model is developed along with erosion filters to improve the traditional atmospheric light factors. A similar erosion operator is also utilized for transmission to refine, and the model is combined with a multi-scale clipping limit of ICLAHE to obtain an optimal nighttime dehazing effect. The experimental results confirm that the proposed strategy performs better in comparison to state-of-the-art techniques and is computationally less intensive.

This paper presents a simple technique for improving the quality of the halftoning-based block truncation coding (H-BTC) decoded image. The H-BTC is an image compression technique inspired from typical block truncation coding (BTC). The H-BTC yields a better decoded image compared to that of the classical BTC scheme under human visual observation. However, the impulsive noise commonly appears on the H-BTC decoded image. It induces an unpleasant feeling while one observes this decoded image. Thus, the proposed method presented in this paper aims to suppress the occurring impulsive noise by exploiting a deep learning approach. This process can be regarded as an ill-posed inverse imaging problem, in which the solution candidates of a given problem can be extremely huge and undetermined. The proposed method utilizes the convolutional neural networks (CNN) and residual learning frameworks to solve the aforementioned problem. These frameworks effectively reduce the impulsive noise occurrence, and at the same time, it improves the quality of H-BTC decoded images. The experimental results show the effectiveness of the proposed method in terms of subjective and objective measurements.

In recent years, the development of document segmentation technology is gaining more and more attention in the area of semantic segmentation, which plays an important role in the task of understanding the structure of documents. Although this demand applying deep learning approaches has undergone continuing advancement, the document segmentation systems still suffer from low accuracy rate. This paper presents a new high-performance document segmentation algorithm, namely Deep Document Segmentation Network (DDSnet), which incorporates advanced end-to-end deep learning methods for segmenting four different types of document features, including background, texts, tables, and figures. To overcome the small receptive fields, the atrous residual block is proposed, which is efficiently boosted by the adoption of multi-branches structure. For the better fine-grained output, the proposed atrous convolution residual block is conducted to achieve high accuracy. Moreover, this paper also releases the brand-new large-scale database, namely PPSD2019, for document segmentation that provides a pixel-level database for another benchmark. As documented in the experimental results, the proposed document segmentation method achieves a superior segmentation rate than that of the former competitive schemes. As a result, the proposed method and database can be considered as a very competitive candidate for the document segmentation applications.

This paper presents a novel deep learning-based technique for reconstructing the Ordered Dithering Block Truncation Coding (ODBTC) decoded image. The proposed method inherits the effectiveness of wavelets transform and Convolutional Neural Networks (CNN). The proposed technique employs the two-dimensional Decimated Wavelet Transform (DWT) to decompose the ODBTC image into low and high frequency sub-bands over various resolution. These image sub-bands are progressively reconstructed with the Pyramid M-Shaped CNN consisting multiple input and output. This scheme considers the ODBTC decoded image as noisy image in Which the information of image can reconstructed by modifying the wavelets sub-bands using residual learning strategy. The produced sub-bands can be combined and transformed back to improve the quality of ODBTC reconstructed image. As documented in experimental results, the proposed method yields promising result for the ODBTC image reconstruction.

Inverse halftoning aims to reconstructs the continuous-tone image, i.e. 255 discrete color levels from its halftone image. The main challenge of inverse halftoning task is on removing the noisy dots over flat areas and, at the same time, restoring the image structures and texture details. To tackle this problem, we propose a novel approach to reconstruct the halftone image in wavelet domain, namely Subbands Aware Inverse Halftoning Network (SAIHN). The proposed framework consists of two sub networks, i.e. the first networks adjust the wavelet sub-bands for image reconstruction, whereas, the second networks refines the quality of reconstructed image by correcting the distortion in pixel domain. The experimental results confirm that the proposed method can reconstruct the halftone image with high quality measured in term of objectively and subjectively assessments. Furthermore, the proposed method demonstrates its superiority on less learnable parameters and more efficient computational complexity in comparison to that of the former existing deep learning-based reconstruction method.

Recent years, fingerprint recognition has a high market share in biometrics, but humans' hands are often covered with oil and sweat that they secrete, rainwater or dirt, which often affect the accuracy of fingerprint recognition. This work proposed a finger vein identification technology for low response time that can be implemented in a cost-efficient embedded system using the binary robust invariant elementary feature. In this method allows the features matching with binary robust invariant elementary feature and verification with multi-quality assessment process. In experimental results, that the EER performance are 0.13% and 0.69%, using homemade and public (FVUSM) datasets when the data were collected with training and testing. The method is very suitable for real-time finger vein recognition applications.

This paper presents a deep learning-based method for effective image companding. The autoencoder inherits the effectiveness of Convolutional Neural Networks (CNN) and residual learning framework to transform High Dynamic Range (HDR) images to Low Dynamic Range (LDR) and its reverse process. Since, the image companding task involves the nondifferentiable operation, thus the encoder and decoder networks are alternately trained using an iterative approach. The experimental result clearly reveals that the proposed method yields a promising result.

This paper proposes a new simple image encryption technique using the Beta chaotic map for performing the confusion and diffusion of input plain image. This image encryption is further applied to the multiple secret sharing (MSS). The former existing scheme in MSS employs the Chinese remainder theorem (CRT) and eXclusive- OR (XOR) operations to convert the secret image to become a set of shared images. The former scheme yields a good result while the number of secret images is even. However, it has slight limitation while the number of secret images is odd. This paper aims to overcome this problem by utilizing the proposed image encryption along with three new MSS approaches: 1) incorporating an additional random images; 2) performing $k$ image encryption; and 3) using two different masking coefficients. As documented in the experimental section, the proposed method performs well in the MSS task. It solves the problem in former scheme on dealing with odd number of secret images. In addition, the proposed method is superior compared with the other competing schemes. The proposed image encryption also improves the security level in the MSS system.

This paper presents an effective image retrieval method by combining high-level features from convolutional neural network (CNN) model and low-level features from dot-diffused block truncation coding (DDBTC). The low-level features, e.g., texture and color, are constructed by vector quantization -indexed histogram from DDBTC bitmap, maximum, and minimum quantizers. Conversely, high-level features from CNN can effectively capture human perception. With the fusion of the DDBTC and CNN features, the extended deep learning two-layer codebook features is generated using the proposed two-layer codebook, dimension reduction, and similarity reweighting to improve the overall retrieval rate. Two metrics, average precision rate and average recall rate (ARR), are employed to examine various data sets. As documented in the experimental results, the proposed schemes can achieve superior performance compared with the state-of-the-art methods with either low-or high-level features in terms of the retrieval rate. Thus, it can be a strong candidate for various image retrieval related applications.

Ocular recognition is expected to provide a higher flexibility in handling practical applications as oppose to the iris recognition, which only works for the ideal open-eye case. However, the accuracy of the recent efforts is still far from satisfactory at uncontrollable conditions, such as eye blinking which implies any poses of eyes. To address these issues, the skin texture, eyelids, and additional geometrical features are employed. In addition, to achieve higher accuracy, sequential forward floating selection is utilized to select the best feature combinations. Finally, the non-linear support vector machine is applied for identification purpose. Experimental results demonstrate that the proposed algorithm achieves the best accuracy for both open eye and blinking eye scenarios. As a result, it offers greater flexibility for the prospective subjects during recognition as well as higher reliability for security.

Recent studies on category relations have shown the promising progress in addressing classification problems. Existing works independently consider the known relation and classifier optimization, and thus restrain the room for performance improvement. In this work, a new loss function is proposed to leverage the underlining relations among categories and classifiers. In addition, the bipolar relation (BR) graph is employed to formulate a general form for diverse relations. This bipolar graph is automatically learnt for reliving the constraints which may happen during the cost minimization. Extensive experiments on three benchmarks with various hypotheses and graphs demonstrate that our method can offer a significant performance improvement by jointly learning from both BR graph and hypothesis, in particular on a small training dataset scenario that suffers from severe overfitting problem.

A contrast enhancement method termed stratified parametric-oriented histogram equalization (SPOHE) is proposed to effectively provide a regional enhanced effect without visual artifacts, e.g., halo or blocking artifacts, which is normally incurred in the former simplified enhancement methods. First, the stratified sampling theory is applied to uniformly sample the original image through many divided strata with the size defined by the two parameters ( $\alpha $ , $\beta $ ). Second, the required statistical information is efficiently derived through the integral image concept. Finally, the corrected SPOHE is also proposed to further improve the contrast with limited tradeoff computations. The experimental results demonstrate that the proposed scheme yields a cumulative distribution function similar to the actual one for an accurate contrast enhancement performance while significantly reducing the computational complexity. Moreover, compared with the former speed-oriented methods, good contrast and artifact-free results can be achieved simultaneously.

Nowadays, the computerized electrocardiogram (ECG) interpretation is the best available tool for the detection of heart diseases. The symptoms of atrial fibrillation, one of the heart disease, are the most challenging task of heart disease that relies on the cardiologist to read the ECG. However, this task involves a time-consuming process, leading to fatigue-induced medical errors. In this paper, a novel atrial fibrillation diagnostic algorithm based on deep learning architecture is proposed which incorporates the spectrogram to further improve the accuracy performance. As shown in the experimental results, the proposed method on PhysioNet /CinC Challenge 2017, which contains 8528 of single-lead ECG recordings, achieves the 10-fold cross-validation set performance of 78%.

Multitone Block Truncation Coding (MT-BTC) image is the superior version of halftone based BTC compression methods. The MT-BTC images are developed based on ordered dithering method which utilize multitone dithering mask for image construction. During the transformation, the original image is processed in a block-wise manner and is replaced in terms of the maximum, minimum and their intermediate values of the respective block. In comparison with standard compressions techniques, MT-BTC possess very unique representation and suffers from inherent halftone noises. In this paper, a simplified version of image to image translation architecture is developed based on cGAN's. To begin with, a MT-BTC database is developed using the latest multitone approach, and it comprise of around 10,000 images. Further, the proposed cGAN's model is optimized to perform with minimal layers and reduced parameters. The PatchGAN discriminator is adjusted to judge over the patch size of 64×64 which has good impact over quality improvements. From the comprehensive performance evaluation, it has been validated that the proposed approach can achieve consistent and improved reconstruction quality.

This paper presents a new technique for improving the quality of Error Difussion Block Truncation Coding (EDBTC) decoded image. The proposed method exploits the usability of wavelet transform on decomposing the EDBTC decoded image into low and high frequency subbands. This new technique employs the two dimensional Discrete Wavelet Transform (DWT) and two dimensional Stationary Wavelet Transform (SWT). This scheme considers the EDBTC decoded image as noisy image in which the occurence of noise can be minimized by modifying the wavelet high frequency. As documented in experimental results, the proposed method yields promising result for the EDBTC image reconstruction.

This paper presents a new simple and efficient technique for secret sharing. This technique inherits the effectiveness of Progressive Visual Secret Sharing (PVSS), in which the quality of recovered secret image is improved if more shared images are stacked together using the eXclusive-OR (XOR) operation. The proposed method also extends the usability of PVSS approach into the Friendly Visual Secret Sharing (PVSS). The content of shared image is converted into more friendly appearances rather than in the noise-like shared image. The proposed method offers promising result for grayscale and color image. It also produces recovered secret image in the lossless condition. The proposed method can be regarded as a good candidate for secret sharing.

Image segmentation is one of the popular techniques for vision application, i.e., retrieving object information from an image. Segmentation not only provides the class and location of an object, but the associated contour as well. Recent advances of the segmentation deploy the encoder and decoder architectures with skip connection, and also utilize the edge information to obtain the best contour. This work focuses on maintaining the information flow from skip connection, and also identifies suited feature from the bottom layer. The feature selector can be also deployed in edge information for yielding segmentation contour. Designed filters are also employed to analyze the best feature for the reconstruction purpose. Experiment results show that the proposed scheme can improvise the performance of the simple ENet to achieve a superior IoU from 51.3% to 57.9%.

Facial expression recognition (FER) is an important and challenging problem for automatic inspection of surveillance videos. In recent years, with the progress of hardware and the evolution of deep learning technology, it is possible to change the way of tackling facial expression recognition. In this paper, we propose a sequence-based facial expression recognition framework for differentiating facial expression. The proposed framework is extended to a frame-to-sequence approach by exploiting temporal information with gated recurrent units. In addition, facial landmark points and facial action unit are also used as input features to train our network which can represent facial regions and its components effectively. Based on this, we build a robust facial expression system and is evaluated using two publicly available databases. The experimental results show that despite the uncontrolled factors in the videos, the proposed deep learning-based solution is consistent in achieving promising performance compared to that of the former schemes.

Digital multitoning is an extension of halftoning for rendering more than two tones at each pixel for higher image quality. Although a lot of effort has been put in generating dispersed dots previously, the blue-noise feature can hardly be achieved for those printers utilizing the electrophotography (EP) process to avoid the physically unstable isolated dots. To overcome this issue, Chandu et al. proposed a screening method for yielding green-noise dot clusters, yet noisy multitone texture was accompanied. This degrades the visual quality and the stability of tone rendering. In this paper, a significantly improved homogeneity of clustered dots can be achieved by the proposed screening method based upon the new inter-iterative clustered-dot direct multi-bit search algorithm. Compared with the former approaches, the inter-iteration design leads to less error by the updated initial multitone patterns. As demonstrated in the experimental results, both of the high homogenous multitone texture and less noisy perception at all absorptance levels are offered in contrast to the former Chandu et al. ’s results. The high-quality output proves it as a very competitive candidate for EP printers, e.g., laser printers.

We propose an effective method to boost the accuracy of multi-person pose estimation in images. Initially, the three-layer fractal network was constructed to regress multi-person joints location heatmap that can help to enhance an image region with receptive field and capture more joints local-contextual feature information, thereby producing keypoints heatmap intermediate prediction to optimize human body joints regression results. Subsequently, the hierarchical bi-directional inference algorithm was proposed to calculate the degree of relatedness (call it Kinship) for adjacent joints, and it combines the Kinship between adjacent joints with the spatial constraints, which we refer to as joints kinship pattern matching mechanism, to determine the best matched joints pair. We iterate the above-mentioned joints matching process layer by layer until all joints are assigned to a corresponding individual. Comprehensive experiments demonstrate that the proposed approach outperforms the state-of-the-art schemes and achieves about 1% and 0.6% increase in mAP on MPII multi-person subset and MSCOCO 2016 keypoints challenge.

Digital halftoning is a strategy to convert the color or gray scale image into a printable format. Among several halftoning techniques, error diffusion is one of the conventional and widely adopted method in many printing devices. In further, halftone classification is very important to obtain the perfect reconstruction of binary printed images. The paper attempts to solve this issue by exploiting the intrinsic similarity between the stochastic geometry and halftoning. Feature vectors are constructed using the point process statistic parameters such as directional distribution function and radially averaged power spectral density. Extreme learning machine model is developed and eight varieties of error diffusion halftone images are considered for classification. From the results, it has been validated that the proposed scheme yields better accuracy of 97.5% and it is faster than the existing approaches.

Error diffusion is an efficient halftone method for mainly being applied on printers. The promising high image quality and processing efficiency endorse it as a popular and competitive candidate in halftoning and multitoning applications. The multitoning is an extension of halftoning, adopting more than two-tone levels for the improvement of the similarity between an original image and the converted image. Yet, the banding effect, indicating the areas with discontinuous tone level, disturbs the visual perception, and thus seriously degrades image quality. To solve the banding effect, the tone-replacement strategy is proposed in this paper. As documented in the experimental results, excellent tone-similarity as that of the original image and promising reconstructed dot-distribution can be provided simultaneously. Comparing with the former banding-free methods, the apparent improvements/features suggest that the proposed method can be a very competitive candidate for multitoning applications.

Compared to the error diffusion, dot diffusion provides an additional pixel-level parallelism for digital halftoning. However, even though its periodic and blocking artifacts had been eased by previous works, it was still far from satisfactory in terms of the blue noise spectrum perspective. In this work, we strengthen the relationship among the pixel locations of the same processing order by an iterative halftoning method, and the results demonstrate a significant improvement. Moreover, a new approach of deriving the averaged power spectrum density (APSD) is proposed to avoid the regular sampling of the well-known Bartlett's procedure which inaccurately presents the halftone periodicity of certain halftoning techniques with parallelism. As a result, the proposed dot diffusion is substantially superior to the state-of-the-art parallel halftoning methods in terms of visual quality and artifact-free property, and competitive runtime to the theoretical fastest ordered dithering is offered simultaneously., Comment: Accepted to IEEE Trans. on Image Processing

This paper presents a new feature descriptor for vehicle verification. The object detection scheme generates the vehicle hypothesis (candidate) that requires subsequent confirmation in the vehicle verification stage with specific feature descriptors. In the procedure of vehicle verification, an image descriptor is generated from the statistical parameter of the curvelet-transformed (CT) subbands. The marginal distribution of CT output is a heavy-tailed bell-shaped function, which can be approximated as Gaussian, Laplace, and generalized Gaussian distribution (GGD) with high accuracy. The maximum likelihood estimation (MLE) produces the distribution parameters of each CT subband for the generation of the image feature descriptor. The classifier then assigns a class label for the vehicle hypothesis based on this descriptor information. As documented in the experimental results, this feature descriptor is effective and outperforms the existing methods in the vehicle verification tasks.

This paper presents a new approach to index color images using the features extracted from the error diffusion block truncation coding (EDBTC). The EDBTC produces two color quantizers and a bitmap image, which are further processed using vector quantization (VQ) to generate the image feature descriptor. Herein two features are introduced, namely, color histogram feature (CHF) and bit pattern histogram feature (BHF), to measure the similarity between a query image and the target image in database. The CHF and BHF are computed from the VQ-indexed color quantizer and VQ-indexed bitmap image, respectively. The distance computed from CHF and BHF can be utilized to measure the similarity between two images. As documented in the experimental result, the proposed indexing method outperforms the former block truncation coding based image indexing and the other existing image retrieval schemes with natural and textural data sets. Thus, the proposed EDBTC is not only examined with good capability for image compression but also offers an effective way to index images for the content-based image retrieval system.

The conventional dual-mode halftoning methods achieve high quality halftone patterns by easing smooth artifact and preserving image details. However, the boundaries between low- and high-frequency image regions still present undesired texture in some particular cases, which significantly degrades the visual quality. In this work, a multi-mode halftoning method is proposed to deal with the object map artifact and boundary artifact simultaneously. In addition, the new absorptance-frequency stacking constraint is also employed to solve the noisy textures of the halftone outputs. As documented in the experimental results, high quality halftone outputs can be obtained, proving that the proposed method can be a competitive candidate for electrophotographic printers.

Degradation in visibility is often introduced to images captured in poor weather conditions, such as fog or haze. To overcome this problem, conventional approaches focus mainly on the enhancement of the overall image contrast. However, because of the unspecified light-source distribution or unsuitable mathematical constraints of the cost functions, it is often difficult to achieve quality results. In this paper, a fusion-based transmission estimation method is introduced to adaptively combine two different transmission models. Specifically, the new fusion weighting scheme and the atmospheric light computed from the Gaussian-based dark channel method improve the estimation of the locations of the light sources. To reduce the flickering effect introduced during the process of frame-based dehazing, a flicker-free module is formulated to alleviate the impacts. The systematic assessments show that this approach is capable of achieving superior defogging and dehazing performance, compared with superior defogging and dehazing performance, compared with the state-of-the-art methods, both quantitatively and qualitatively.

Formerly, dimensionality reduction techniques are effective ways for extracting statistical significance of features from their original dimensions. However, the dimensionality reduction also induces an additional complexity burden which may encumber the real efficiency. In this paper, a technique is proposed for the reduction of the dimension of samples rather than the features in the former schemes, and it is able to additionally reduce the computational complexity of the applied systems during the reduction process. This method effectively reduces the redundancies of a sample, in particular for those objects which possess partially symmetric property, such as human face, pedestrian, and license plate. As demonstrated in the experiments, based upon the premises of faster speeds in training and detection by a factor of 4.06 and 1.24, respectively, similar accuracies to the ones without considering the proposed method are achieved. The performance verifies that the proposed technique can offer competitive practical values in pattern recognition related fields.

This study analyzes the recent image watermarking schemes based on redundant discrete wavelet transform (RDWT) and singular value decomposition (SVD), and shows that in fact they are insecure and cannot be used for protecting the rightful ownership. The RDWT-SVD watermarking directly embeds a grayscale watermark image of the same size with the host image into the singular value matrix of the RDWT-transformed host image, then produces the left and right orthogonal matrices as side information which is later used in the watermark extraction stage. The RDWT-SVD approach enjoys the advantage of the RDWT redundancy to achieve a high embedding capacity, and preserves the watermark imperceptibility by exploiting the SVD stability properties. It is claimed that RDWT-SVD watermarking is robust against several common image processing and geometrical attacks, yet a fundamental flaw in the RDWT-SVD scheme is found, which leads to severe the false positive issue. Three vulnerable attacks should be considered in the RDWT-SVD scheme: (1) An attacker can easily claim the owner watermarked image; (2) the owner has the ambiguity because of the wrong side information usage, and (3) the owner can extract the correct watermark from arbitrary image. Thus, it is important to highlight these attacks when implementing the RDWT-SVD watermarking scheme.

Moving object detection is an important and fundamental step for intelligent video surveillance systems because it provides a focus of attention for post-processing. A multilayer codebook-based background subtraction (MCBS) model is proposed for video sequences to detect moving objects. Combining the multilayer block-based strategy and the adaptive feature extraction from blocks of various sizes, the proposed method can remove most of the nonstationary (dynamic) background and significantly increase the processing efficiency. Moreover, the pixel-based classification is adopted for refining the results from the block-based background subtraction, which can further classify pixels as foreground, shadows, and highlights. As a result, the proposed scheme can provide a high precision and efficient processing speed to meet the requirements of real-time moving object detection.

This work presents a novel reversible data-hiding scheme which embeds secret data into a side matched Vector Quantization (SMVQ)-compressed image, and achieves lossless reconstruction of a Vector Quantization (VQ)-compressed image. The rather random distributed histogram of a VQ-compressed image can be re-located to locations close to zero by SMVQ prediction. Thus, fewer bits can be used to encode SMVQ indices with very small values, and no indicator is required to encode these indices, which yields extra hiding space to hide secret data. Consequently, high embedding capacity and low bit rate scenarios can be deposited. Experimental results demonstrate the effectiveness and reversibility of the proposed scheme. Moreover, in terms of the embedding rate, the bit rate, and the embedding capacity, experimental results show that the performance of the proposed scheme is better than those of the former data hiding schemes for VQ-based and VQ/SMVQ-based compressed images.

Block truncation coding (BTC) has been considered as a highly efficient compression technique for decades, but the blocking artifact is its main issue. The halftoning-based BTC has significantly eased this issue, yet an apparent impulse noise artifact is accompanied. In this study, an improved BTC, termed adaptive dot-diffused BTC (ADBTC), is proposed to further improve the visual quality. Also, this method provides an additional flexibility on the compression ratios determination in contrast to the former fixed and few number of configuration possibilities. As documented in the experimental results, the proposed method achieves the superior image quality regarding the five various objective IQA methods. As a result, it is a very competitive approach for the needs of both high frame rate and high-resolution image compression.

Block truncation coding (BTC) is an efficient compression technique with extremely low computational complexity. However, the blocking and false contour effects are two major deficiencies in BTC which cause severe perceptual artifacts. The former scheme, error-diffused BTC (EDBTC), can significantly improve the above issues through the visual low-pass compensation on the bitmap, which thus widens its possible application market, yet the corresponding security issue may limit its value. In this paper, a method namely complementary hiding EDBTC is developed to cope the above issue. This paper is managed by firstly discussing when a single watermark is embedded, and then multiple watermarks are employed to test the limitation of the proposed scheme. Herein, an adaptive external bias factor is employed to control the watermark embedding, and this factor also affects the image quality and robustness simultaneously. Experimental results demonstrate that the proposed method only requires an extremely small external bias factor to carry watermarks, which enables a high capacity scenario without significantly damaging image quality.

In this paper, the impact of the lips for identity recognition is investigated. In fact, it is a challenging issue for identity recognition solely by the lips. In the first stage of the proposed system, a fast box filtering is proposed to generate a noise-free source with high processing efficiency. Afterward, five various mouth corners are detected through the proposed system, in which it is also able to resist shadow, beard, and rotation problems. For the feature extraction, two geometric ratios and ten parabolic-related parameters are adopted for further recognition through the support vector machine. Experimental results demonstrate that, when the number of subjects is fewer or equal to 29, the correct accept rate (CAR) is greater than 98%, and the false accept rate (FAR) is smaller than 0.066%. (CAR95.02%, FAR0.095% # Subjects ≤ 57). Moreover, the processing speed of the overall system achieves 34.43 frames per second, which meets the real-time requirement. Thus, the proposed system can be an effective candidate for facial biometrics applications when other facial organs are covered or when it is applied for an access control system.

This paper presents an improved hand tracking system using pixel-based hierarchical-feature AdaBoosting (PBHFA), skin color segmentation, and codebook (CB) background cancellation. The proposed PBH feature significantly reduces the training time by a factor of at least 1440 compared to the traditional Haar-like feature. Moreover, lower computation and high tracking accuracy are also provided simultaneously. Yet, one of the disadvantages of the PBHFA is the false positive which is the consequence of the appearance of complex background in positive samples. To effectively reduce the false positive rate, the skin color segmentation and the foreground detection by applying the CB model are catered for rejecting all of the candidates which are not hand targets. As documented in the experimental results, the proposed system can achieve promising results, and thus it can be considered as an effective candidate in handling practical applications which require hand postures.