Mobile QR Code QR CODE

A Practical Light Field Representation and Coding Scheme with an Emphasis on Refocu

(Chun Zhao) ; (Byeungwoo Jeon)

Light field images have drawn a great deal of interest owing to their flexible postprocessing and versatile functionalities. However, the tremendous data volume has imposed a practical limit in applications, which requires a better way to represent and compress the data. In this paper, we propose a practical representation and coding scheme for light field data with a special emphasis on retaining the refocusing functionality as much as possible. Under the proposed method, light field data are represented and compressed in the form of a rendered all-in-focus image and a depth map. Before encoding, the all-in-focus image is rendered accurately based on the depth map using the proposed difference focus measure. At the decoder, a focal stack comprising multiple images having different focus levels can be reconstructed by applying a defocusing function to the compressed all-in-focus image where the defocusing function parameter depends on the selected depth level of the compressed depth map. Compared with well-known state-of-the-art methods that compress the Fourier Disparity Layer or the focal stack, the proposed coding scheme shows much smaller loss in the refocusing capability (16.2% and 17.8% less than two state-of-theart methods) and provides PSNR improvements of 1.60dB to 2.38dB at the same compression ratio.

Information Extraction from Invoices by using a Graph Convolutional Neural Network: A Case Study of Vietnamese Stores

(An Cong Tran) ; (Lai Thi Ho) ; (Hai Thanh Nguyen)

Information extraction automatically obtains structured information from unstructured or semi-structured machine-readable documents. The extraction steps consist mainly of classifying words (tagging). The output can be stored as key-value pairs in a computer-friendly file format, and then stored in a database for later reference. Information extraction from receipts or invoices is a difficult task because the tagging step should not be done solely on machine-readable words. Also, we obtain layout information or positions of words relative to other words in the invoices or receipts. This study deployed optical character recognition solutions for the Vietnamese language (VietOCR) combining a graph convolutional network (GCN) to extract information from 731 Vietnamese invoices issued by several stores. First, we collected invoice images captured with smartphones from supermarkets in Vietnam. Then, with those images we proceeded with text detection and recognition, then feature processing. The dataset was classified into two parts for training and testing, and we executed classification tasks with two GCNs. Experimental results revealed that our proposed method reached 99.50%, 98.52%, 98.52%, and 98.52% for accuracy, recall, precision, and F1-score, respectively. This work is expected to prove useful for information extraction from image-based documents.

Learning-based Framework for Color Conversion between Digital Cinema Package and Streaming Movie

(Younghoon Lim) ; (Kwanwoo Park) ; (Joonki Paik)

Humans feel different experiences when viewing a cinema at the theater and home. This paper presents methods to reduce the difference in the viewing experience. Based on the workflow of digital cinema distribution, the proposed method attempts to minimize the color difference between digital cinema packages (DCP) for the theater and streaming movies. For end-to-end mapping between the DCP and streaming movie, this paper proposes a convolutional neural network (CNN)-based color conversion algorithm based on the SMPTE standard. The proposed method consists of three steps: i) color conversion using standard matrices, ii) color conversion using the CNN, and iii) color saturation error removal by fusing the results in steps i) and ii). The proposed method enhances the color of TV streaming images because it minimizes the color difference from the DCP and appropriately extends the color gamut. As a result, the proposed method can provide consumers with indistinguishable quality from a DCP movie at the theater.

A Painting Style System using an Improved CNN Algorithm

(Yuan Zhong) ; (Xinyan Huang)

The rapid development of deep learning technology allows ordinary people to create artwork that imitates the style of paintings by famous masters through an algorithm. To create such works with artistic style, this research proposes an artificial neural network algorithm based on an improved convolutional neural network (CNN). First, a fast style-rendering model based on the improved CNN is constructed, and then, a server front end is built with the Bootstrap framework. The server-side back end of the system is built by combining a Python algorithm and a web framework, and finally, a complete model of the front-end and back-end network of the style rendering system is constructed. The model proposed in this paper is compared with two other models to verify its performance. The results show that information entropy of the model constructed is the highest at 5.58, which is higher than information entropy of the other two models. The average gradient value and the peak signal-to-noise ratio under the constructed model are 22.54 and 27.81, respectively, which are also higher than the other two models. Mutual information and the structural similarity index between rendered images and sample images under all three models were compared. Mutual information and structural similarity index of the model constructed by this research are 1.19 and 0.56, respectively, with much larger data sizes than the two comparison models.

Progressive Dehazing and Depth Estimation from a Single Hazy Image

(Jeonghoon Kim) ; (Sungyoon Kim) ; (Changhoon Pyo) ; (Hyeongmyeon Kim) ; (Changhoon Yim)

We propose a progressive dehazing and depth estimation (PDDE) method with optimal estimation of the attenuation coefficient. We investigated the characteristics of dehazing operations related to the depth and attenuation coefficient for PDDE. Entropy was used as a non-reference image quality metric for optimality assessment of dehazed images. The proposed PDDE method provides an optimal dehazed image and depth estimation from a single hazy image. Experimental results show that the proposed method provides clearer and subjectively better results for single image dehazing than previous image dehazing methods. It also results in a significant improvement in the accuracy of depth estimation from a single hazy image compared with previous depth estimation methods.

Cryptography for Privacy in a Resource-Constrained IoT: A Systematic Literature Review

(Fredy Mendoza-Cardenas) ; (res J. Aparcana-Tasayco) ; (Rai Stiv Leon-Aguilar) ; (Jose Luis Quiroz-Arroyo)

The Internet of Things (IoT) is intelligent devices connected to a network, which enables collection, transmission, and processing of real-world data. IoT applications make extensive use of resource-constrained devices. These applications work with critical data that need confidentiality and privacy during transmission. In this resource-constrained scenario, lightweight cryptography is a tool to ensure confidentiality, privacy, and data integrity. The number of research papers on cryptography in a resource-constrained IoT has increased over three years. A systematic literature review (SLR) is a search for papers and answers on relevant questions for researchers in their area of interest. This study applies an SLR based on the Kitchenham and PRISMA guidelines to analyze the influence of cryptography on privacy in a resource-constrained IoT. The most relevant results show that the number of papers published on the topic in the years 2019-2021 has increased, that most of the papers found in the review were in IEEE Access, and that the most frequent words in the titles and abstracts were lightweight, IoT, encryption, and scheme. Moreover, most of the research was developed on the Asian continent. This study concludes that cryptography for privacy in a resource-constrained IoT is a developing area of research.

Network Monitoring System for Software-defined Networking: Experimental Evaluation in a Real SDN Environment

(Fredy Mendoza-Cardenas) ; (res J. Aparcana-Tasayco) ; (Daniel Diaz-Ataucuri)

Software-Defined Networking (SDN) is a data network paradigm that defines the separation of the data plane and control plane so that applications can change network behavior through software algorithms. Many academic solutions are focused on network performance reduction. However, these works lack experimental evaluation in a physical environment. Many of these proposed solutions were evaluated in simulated/emulated environments, so these solutions are not suitable for production cases. The present research determined the influence of a proposed Networking Monitoring System (NMS) in a real SDN environment. Relevant results and an exhaustive statistical test were done to evaluate the influence of NMS on the physical server performance in a real SDN environment. The average CPU usage with ONOS is 1.47%, but the average CPU usage with ONOS + NMS is higher at 1.84%. Memory usage average with ONOS is 18.81 GB, but the average memory usage with ONOS + NMS is higher at 19.24 GB. The focus of NMS is on being a platform that permits future networking specialists to learn and research in a real SDN environment. Future networking researchers are encouraged to use production controllers such as ONOS, which works with OF-DPA and Pica8 switches.

Hample filter based Short term Solar Forecasting using Facebook Prophet Library in Energy Harvested Wireless Sensor Network (EHWSN)

(Vivek Kumar Verma) ; (Satya Sai Srikant)

Energy harvesting prolongs the life of wireless sensor networks. Among all ambient sources of energy, solar energy is believed to be the most efficient in terms of energy density but fluctuates with time and location. The main objective of this study is to forecast solar energy for Energy Harvested Wireless Sensor Network (EHWSN) using the Facebook Prophet Library with and without a filter and to assess the model efficacy. The present study examined the models using National Renewal Energy Laboratory (NREL) one-minute sample data for June 2010 to 2016. The outliers were eliminated using a Hample filter with a standard deviation of 3 and a window size of 3. The forecast error was measured using the cross-validation feature of the Prophet Library. The Mean squared Error ( MSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and Root Mean Squared Error (RMSE ) values for the Prophet model were 292.81, 14.56, 19.37, and 17.11, respectively, while they were 92.56, 14.86, 14.86, and 17.10, respectively, for the Prophet model based on the Hample Filter. MAPE was reduced by 18.60 %, while the rest of the metrics changed significantly. The Prophet and Hample?Prophet give the best result regarding RMSE and accuracy compared with previous work based on the Ensemble Approach.

FFDSOGI-PLL-based DSTATCOM for Power Quality Enhancement

(Balasubramanian M) ; (Nagarajan C)

This paper proposes a control technique based on a frequency-fixed dual second-order generalized integrator phase-locked loop (FFDSOGI-PLL) for a Distribution Static Compensator (DSTATCOM). The goal is to mitigate power-quality (PQ) issues in a three-phase distribution network with a weak grid. The phase angle, symmetrical components, and positive sequence of supply voltages were estimated using the FFDSOGI-PLL even in a polluted grid environment. The active component magnitude of the fundamental load current was extracted using SOGI, which also reduces the PQ issues by generating compensating current. The effectiveness of the proposed FFDSOGI-PLL-based DSTATCOM was verified for functions like load balancing, power factor correction (PFC), harmonics reduction, etc. This was done in MATLAB/SIMULINK under steadystate and dynamic conditions. The simulation results show that the FFDSOGI-PLL-based control algorithm has faster convergence and quick dynamic response. Hence, the DSTATCOM performance is satisfactory for mitigating PQ issues with the FFDSOGI-PLL-based control technique.

A High-Throughput and Energy-Efficient SHA-256 Design using Approximate Arithmetic

(Junhyuk Baik) ; (Yongtae Kim)

This paper presents a novel high-throughput SHA-256 design exploiting approximate computing. The 32-bit addition of the SHA-256 architecture consumes a lot of hardware resources, hence we suggest a split k-bit adder-based SHA-256 design to significantly reduce hardware resources while ensuring acceptable hash functionality. The proposed designs enhance the area, delay, and energy when employed with 28-nm CMOS technology by 8.9%, 45.7%, and 9.3% in the 2-bit configuration and 15%, 52.9%, and 11.3% in the 1-bit configuration, respectively, compared to the conventional SHA-256 design. Additionally, the proposed designs improve the area-delay product (ADP), energy-delay product (EDP), and throughput performances by at least 19.1%, 20.2%, and 18.6% with a maximum of 149.8%, 139.3%, and 112.1%, respectively, compared to the traditional design. Additionally, all of the proposed designs obtain a higher average of Avalanche effect than the traditional ones, and all the message digests generated by the proposed SHA-256 are randomly distributed.