Search

In this study, we propose a fire classification system using image clustering based on a federated learning (FL) structure. This system enables fire detection in various industries, including manufacturing. The accurate classification of fire, smoke, and normal conditions is an important element of fire prevention and response systems in industrial sites. The server in the proposed system extracts data features using a pretrained vision transformer model and clusters the data using the bisecting K-means algorithm to obtain weights. The clients utilize these weights to cluster local data with the K-means algorithm and measure the difference in data distribution using the Kullback–Leibler divergence. Experimental results show that the proposed model achieves nearly 99% accuracy on the server, and the clustering accuracy on the clients remains high. In addition, the normalized mutual information value remains above 0.6 and the silhouette score reaches 0.9 as the rounds progress, indicating improved clustering quality. This study shows that the accuracy of fire classification is enhanced by using FL and clustering techniques and has a high potential for real-time detection.

In the elderly population, falls are one of the leading causes of fatal and non-fatal injuries. Fall detection and early alarms play an important role in mitigating the negative effects of falls, especially given the growing proportion of the elderly population. Due to their non-intrusive nature, data availability, and low deployment costs, RGB videos have been used in many previous studies to detect falls. The RGB data, however, can be affected by background environment changes, resulting in non-recognition. To overcome these challenges, many researchers propose extracting skeleton data from RGB videos and using it for fall detection. Although there have been multiple surveys on fall detection, most of them focus on assessing fall detection systems using different kinds of sensors, and a comprehensive evaluation of skeleton-based fall detection in RGB videos is lacking. In this paper, we examine the most recent advances in skeleton-based fall detection in RGB videos, from handcrafted feature-based methods to advanced deep learning algorithms. Further, we present several skeleton-based fall detection techniques and their performance results on various benchmark datasets, along with challenges and future directions in this field.

In the last decades, deep learning (DL) has emerged as a powerful and dominant technique for solving challenging problems in various fields. Likewise, in the field of digital image forensics, a large and growing body of literature investigates DL-based techniques for detecting and classifying tampered regions in images. This article aims to provides a comprehensive survey of state-of-the-art DL-based methods for image-forgery detection. Copy-move images and spliced images, two of the most popular types of forged images, were considered. Recently, owing to advances in DL, DL-based approaches have yielded much better results as compared to traditional non-DL-based ones. The surveyed techniques were proposed by developing or fusing various efficient DL methods, such as CNN, RCNN, or LSTM to adapt to detecting tampered traces.

Recent studies on image stitching have been extensively conducted to stitch panoramic or 360° images using a small number of input images. The stitching of aerial images, that are captured by unmanned aerial vehicles (UAVs), has various practical applications. In this paper, we propose a fast adaptive stitching algorithm for handling numerous aerial images. First, the proposed method analyzes the relative positions and overlapping regions of the UAV image footprints by exploiting their geotag information. Based on the analysis, an adaptive selection algorithm is proposed to eliminate the densely overlapped images from among all the UAV images. Then, the proposed method sequentially performs fast feature extraction and feature matching. Finally, a local warp method, with a smooth transition for overlapping regions, is introduced to alleviate the blurring artifacts and achieve highly accurate image alignment. The experiments are conducted for various scenarios to generate seamless terrestrial mosaic images of large areas. The proposed method improves the visual quality of the stitched image, by decreasing the estimated reprojection error and the number of observed visual distortions. In addition, the proposed method can substantially reduce the processing time compared with conventional stitching methods.

In the image forgery problems, previous works has been chiefly designed considering only one of two forgery types: copy-move and splicing. In this paper, we propose a scheme to handle both copy-move and splicing image forgery by concurrently classifying the image forgery types and localizing the forged regions. The structural correlations between images are employed in the forgery clustering algorithm to assemble relevant images into clusters. Then, we search for the matching of image regions inside each cluster to classify and localize tampered images. Comprehensive experiments are conducted on three datasets (MICC-600, GRIP, and CASIA 2) to demonstrate the better performance in forgery classification and localization of the proposed method in comparison with state-of-the-art methods. Further, in copy-move localization, the source and target regions are explicitly specified.

Recently, the task of validating the authenticity of images and the localization of tampered regions has been actively studied. In this paper, we go one step further by providing solid evidence for image manipulation. If a certain image is proved to be the spliced image, we try to retrieve the original authentic images that were used to generate the spliced image. Especially for the image retrieval of spliced images, we propose a hybrid image-retrieval method exploiting Zernike moment and Scale Invariant Feature Transform (SIFT) features. Due to the symmetry and antisymmetry properties of the Zernike moment, the scaling invariant property of SIFT and their common rotation invariant property, the proposed hybrid image-retrieval method is efficient in matching regions with different manipulation operations. Our simulation shows that the proposed method significantly increases the retrieval accuracy of the spliced images.

A spatial and interlayer hybrid intra-prediction (SILHIP) scheme is proposed in this work to improve the coding efficiency of the scalable video coding (SVC) standard. The SVC standard adopts a line-by-line intra-prediction that predicts the texture information of a block by exploiting the spatial correlation between adjacent blocks. However, the accuracy of the line-by-line intra-prediction deteriorates when a target block lies in the edge region. The proposed SILHIP scheme adaptively predicts a target block by exploiting both the spatial correlation and the interlayer correlation between the base layer and the enhancement layer. To implement the SILHIP scheme, we partition a macroblock into two regions based on the edge map and allow different types of prediction schemes to be used in different regions. It is shown by experimental results that the proposed SILHIP scheme outperforms the conventional one consistently. [ABSTRACT FROM AUTHOR]

The scalable video coding (SVC) standard adopts four post-processing error concealment (EC) methods as a non-normative tool in order to conceal a lost picture in transmission. In this paper, we first introduce an error tracking model with which the concealment error and propagation error can be estimated. Then, we analyze the four EC methods using the error tracking model. Until now, there have been few studies on how to conceal SVC bitstream corrupted by burst errors. Based on the results of the error analysis, we propose a selection-based EC strategy to conceal burst errors for SVC bitstream. In the proposed strategy, one of the four EC methods is adaptively selected such that the visual degradation is minimized for each lost picture. The experimental results show that the proposed EC strategy can improve the EC performance without increasing the computational complexity. Moreover, the proposed strategy does not require any modification of the conventional SVC codec since it utilizes only the EC methods adopted in the standard. [ABSTRACT FROM AUTHOR]

Scalable video coding (SVC) is currently developed as an extension of H.264/AVC. In the SVC encoder, an exhaustive search technique is employed to select the best coding mode for each macro block (MB). This technique achieves an optimal trade-off between rate and distortion, but it requires an extremely large encoding time. In this paper, we propose a fast mode decision algorithm for inter-layer coding at the enhancement layer. The proposed algorithm predicts the mode of each MB at the enhancement layer using the modes of a co-located MB and its neighboring MBs at the base layer. The proposed method can achieve a time saving of up to 74 % in spatial scalability and 63 % in coarse grain quality scalability with negligible loss of quality and bit rate increment. [ABSTRACT FROM AUTHOR]

In the wireless microsensor networks, both source and sink nodes can be changed or mobile. The movement of source and sink may lead to the breakage of existing routes. In most routing protocols, query packets are broadcasted to correct a broken route between source and sink, which causes significant communication overhead in terms of both energy and delay. In order to support the sink mobility of conventional routing protocols, we propose a simple route maintaining algorithm which does not use the flooding method. Since the proposed algorithm does not require the information on the geometric location of sensor nodes, it can be easily adopted in most existing routing protocols including Ad Hoc On Demand Distance Vector (AODV) and Direct Diffusion (DD). Experimental results show that the proposed algorithm drastically improves the conventional routing protocols in terms of both energy and delay in case of mobile sinks. [ABSTRACT FROM AUTHOR]

The scalable video coding (SVC) standard adopts the inter-layer residual prediction (ILRP) algorithm to encode the residual signal of the enhancement layer (EL). The ILRP reduces the number of bits required for encoding the residual signal but incurs excessive encoding time. In this paper, we propose a fast encoding algorithm for SVC-based video streaming. In this algorithm, the ILRP is selectively applied to the coding modes depending on their Lagrangian rate-distortion costs. Experimental results show that the proposed algorithm can reduce computational complexity with negligible video quality degradation and bitrate increments. [ABSTRACT FROM AUTHOR]

In many applications including Internet video broadcast and interactive TV, video streams are transmitted over various networks. In order to fulfill the requirement of the users who want to watch a part of the video stream, start playback at a random location, and jump to another location, the service provider has to implement the random access functionality. In addition, when the video stream is transmitted over error-prone networks like wireless LAN, the random access point can be used as a recovery point which enables the user device to refresh the decoding process corrupted by the transmission error. In this paper, we propose a simple transcoding algorithm inserting the random access point in the pre-encoded video stream. The multimedia server adopting the proposed transcoding algorithm can support the error-resilient transmission as well as the playback at random location. Experimental results show that the proposed method can gain significant reduction in the complexity without introducing degradation on visual quality. [ABSTRACT FROM AUTHOR]

Intelligent digital TV (iDTV) is a future digital TV with intelligence which can automatically provide user-personalized services for each audience. For the user-personalized services, the iDTV should recognize audiences in real-time. In this paper, we define a novel structure of the iDTV and propose a real-time person identification system in the iDTV that analyzes captured images and recognizes audiences. The proposed system consists of three processing units: preprocessing for reducing computational costs of the proposed system, face detection using a statistical approach, and face recognition using Support Vector Machines (SVMs). Experimental results show that the proposed system achieves efficient performance with high recognition accuracy of 90% or higher at the speed of 15∼20 fps, which is suitable for the iDTV. [ABSTRACT FROM AUTHOR]

Recently, there is an active research about a new emerging video coding standard, scalable video coding (SVC). SVC adopts a layered coding scheme to generate multi-layered bitstream for heterogeneous environments. In addition to the layered coding, SVC performs temporal prediction using the hierarchical B picture scheme in order to provide temporal scalable representation of input sequences. Since the coding order of the hierarchical B picture scheme is different from that of conventional temporal prediction schemes, the layer-switching algorithm considering the djfferent coding order has to be developed. In this paper, we first investigate the problems of layer switching in SVC and propose an algorithm that enables the SVC decoder to cope with layer switching caused by the bandwidth variation of networks. With the proposed algorithm, the SVC decoder can decode the time-varying sub-bitstream without modification of encoder. We confirm the proposed methods through the experiments. [ABSTRACT FROM AUTHOR]

A macroblock-level deblocking method is proposed for H.264/AVC, in which blocking artifacts are effectively eliminated in the discrete cosine transform domain at the macroblock encoding stage. Experimental results show that the proposed algorithm outperforms conventional H.264 in terms of coding efficiency, and the bitrate saving is up to 5.7% without reconstruction quality loss. [ABSTRACT FROM AUTHOR]

The three-dimensional (3D) extension of the high-efficiency video coding (3D-HEVC) standard adopts new depth-modelling modes (DMMs) to provide an alternative prediction scheme for depth-map intra-coding. In 3D-HEVC, although edges in depth maps can be accurately estimated by utilising the DMMs, testing the DMMs in the mode decision introduces a huge computational load to the encoder. A fast mode decision algorithm is proposed that can significantly reduce the computational overhead incurred by the DMMs. Experimental results indicate that the proposed algorithm reduces the encoding complexity by 33.89% on average without any penalty in coding efficiency. [ABSTRACT FROM AUTHOR]

A transform-domain successive elimination algorithm for block-based motion estimation (ME) is proposed. First, a method is introduced for calculating the sum of squared differences in the transform domain. Then, based on this method, a new transform-domain criterion is proposed to eliminate impossible candidates as early as possible. Experimental results show that the proposed algorithm speeds up the ME process by up to a factor of 10 compared with the conventional full search algorithm. [ABSTRACT FROM AUTHOR]

A novel post-processing technique in the DCT domain, based on the theory of the projection onto convex sets (POCS), to reduce the blocking artifacts in the block discrete cosine transform (BDCT)-coded images, is proposed. Computer simulation results indicate that the proposed scheme performs better than conventional post-processing algorithms. Furthermore, the proposed algorithm can be implemented without DCT=IDCT operations. Estimates of computation savings vary between 41 and 64% depending on the task. [ABSTRACT FROM AUTHOR]

We present an adaptive quarter-pel (Qpel) motion estimation (ME) method for H.264/AVC. Instead of applying Qpel ME to all macroblocks (MBs), the proposed method selectively performs Qpel ME in an MB level. In order to reduce the bit rate, we also propose a motion vector (MV) encoding technique that adaptively selects a different variable length coding (VLC) table according to the accuracy of the MV. Experimental results show that the proposed method can achieve about 3 average bit rate reduction. [ABSTRACT FROM AUTHOR]

We propose a new intra prediction algorithm that can improve the coding efficiency of the H.264/AVC standard. In the proposed algorithm, the prediction signal for each macroblock (MB) is formed by using all the previously reconstructed pixels of the current frame. Experimental results show that the proposed algorithm can reduce the bit rate by 1.36 to 9.07, as compared with the conventional intra prediction, while the average PSNR is not decreased. [ABSTRACT FROM AUTHOR]

Tempcore process simulator (TPS) has been developed in this study to analyze the microstructural evolution of quenched and tempered rebar. There has been an increasing need to relate the complex microstructures to the resulting properties of quenched and tempered rebar. However, information on such relationships typically requires precise thermal histories imposed on the workpiece. Therefore, TPS, capable of simulating the Tempcore process, has been developed to produce high-fidelity data. TPS mainly consists of a vacuum induction furnace, pilot rolling mill, box furnace, and cooling unit to simulate shop floor operations. A series of experimental tests were successfully carried out with various parameters, such as reheating temperature, water flow, water pressure, and cooling time. The effects of chemical compositions and cooling time on the microstructural evolution and mechanical properties of quenched and tempered rebar have been analyzed to validate the performance of TPS. The results show that TPS can simulate the Tempcore process with a high degree of fidelity and reliability.