The areas of optoelectronic systems for image/video acquisition and their applications have expanded rapidly in recent years. Advanced optoelectronic sensors, novel programmable optical modulators, combined with smart computational theories and algorithms, produce novel optoelectronics imaging systems that expand the spatial and temporal dimension for image/video acquisition. Multimedia and data acquired by these new optoelectronic systems impose new challenges in analysis and processing. This conference will provide an international forum for the discussion of the state-of-the-art techniques of optoelectronic system and multimedia processing. Invited talks will be presented by leading scientists in the relevant fields. We are soliciting papers in all relevant areas including, but not limited to, the following: ;
In progress – view active session
Conference 11897

Optoelectronic Imaging and Multimedia Technology VIII

In person: 10 - 11 October 2021
On demand now
View Session ∨
  • Opening Ceremony and Plenary Session I
  • 1: Computer Vision and Graphics
  • Plenary Session II
  • 2: Hyperspectral and Single-Pixel Imaging
  • 3: Biomedical Imaging and Microscopy
  • Poster Session
Opening Ceremony and Plenary Session I
In person / Livestream: 10 October 2021 • 09:00 - 12:00 China Standard Time
9:00: Opening Ceremony
9:20: Awards and Recognition

11900-501
Author(s): Rebecca R. Richards-Kortum, Director, Rice 360 Institute for Global Health Technologies (United States), Rice Univ. (United States)
In person / Livestream: 10 October 2021 • 09:30 - 10:10 China Standard Time
Show Abstract + Hide Abstract
This talk will examine the challenges of designing and translating new low-cost optical technologies to improve global health equity, drawing from examples to improve newborn survival in African hospitals, to improve early detection of cervical cancer for women in Texas and Latin America, and to improve point-of-care diagnosis of COVID-19. The talk will summarize lessons learned to increase the diversity of innovation teams, and to increase the impact and sustainability of the resulting innovations.
11890-502
Author(s): Wentao Wang, State Key Lab. of High Field Laser Physics (China), CAS Ctr. for Excellence in Ultra-intense Laser Science (China), Shanghai Institute of Optics and Fine Mechanics, Chinese Academy of Sciences (China)
In person: 10 October 2021 • 10:40 - 11:20 China Standard Time
Show Abstract + Hide Abstract
X-ray free-electron lasers can generate intense and coherent radiation at wavelengths down to the sub-ångström region and have become indispensable tools for applications in structural biology and chemistry, among other disciplines. Several X-ray free-electron laser facilities are in operation; however, their requirement for large, high-cost, state-of-the-art radio-frequency accelerators has led to great interest in the development of compact and economical accelerators. Laser-wakefield accelerators can sustain accelerating gradients more than three orders of magnitude higher than those of radio-frequency accelerators, and are regarded as an attractive option for driving compact X-ray free-electron lasers. However, the realization of such devices remains a challenge owing to the relatively poor quality of electron beams that are based on a laser-wakefield accelerator. After ten years of efforts, we present an experimental demonstration of undulator radiation amplification in the exponential-gain regime by using electron beams based on a laser-wakefield accelerator.
11893-503
Author(s): Rui Zhu, Peking Univ. (China)
In person: 10 October 2021 • 11:20 - 12:00 China Standard Time
Show Abstract + Hide Abstract
Perovskite solar cells have attracted tremendous attention in recent years due to the high device performance and the superior optoelectronic properties of perovskite materials. In this talk, I will give an introduction about the advances of perovskite solar cells. Then, I will present our efforts on improving the device performance and understanding the device physics of perovskite solar cells. In addition, I will also discuss the advantages of perovskite solar cells for space aircraft application. We have some preliminary efforts of sending the perovskite solar cells into near space. I will share our view about the exciting possibilities for perovskite solar cell technology.
Session 1: Computer Vision and Graphics
In person: 10 October 2021 • 13:30 - 15:10 China Standard Time
Session Chairs: Qionghai Dai, Tsinghua Univ. (China), Zhenrong Zheng, Zhejiang Univ. (China)
11897-2
Author(s): Yanfa Xiang, Qiming Ren, Rui-pin Chen, Zhejiang Sci-Tech Univ. (China)
On demand | Presented Live 10 October 2021
Show Abstract + Hide Abstract
The underwater polarization dehazing imaging has attracted a lot of interest due to the potential applications in corresponding fields. There are some progress on the underwater polarization dehazing imaging by introducing the deep learning into polarization dehazing imaging. In this work, the underwater active polarization dehazing imaging based on the deep learning model is studied. A modified All-in-One Dehazing Network model with three input channels is designed under the framework of TensorFlow. The polarization image data of three different polarization components are designed as the training set with the convolution neural network (CNN).This light-weight CNN is designed to achieve underwater dehazing imaging of different targets with different turbidity. Experiment results indicate that the prediction and estimation using modified AOD-Net have better accuracy than that of the traditional dehazing model.
11897-3
Author(s): Bote Qi, Lihua Shen, Rui-pin Chen, Zhejiang Sci-Tech Univ. (China)
On demand | Presented Live 10 October 2021
Show Abstract + Hide Abstract
Light scattering is inevitable in optical imaging and random speckles will be formed when light passes through the scattering media due to the influence of multiple scattering. Recent research results show that the super-resolution imaging can be achieved by the scattering effect.In this work, the effective focusing of the laser beam through the is realized when the wavefront of the incident light is modulated by the feedback optimization algorithm (area-by-area modulation algorithm) with the spatial light modulator. The experiment results indicate that the maximum light intensity growth factor (132 in this work) and the focusing spot size (1/10 compared to the traditional lens focusing) are dependent on the total number of modulation units on the spatial light modulator and the phase accuracy of each modulation unit. In particular, the focusing of a vector beam through the scattering media provides a new way to control the the maximum light intensity growth factor and the focusing spot size.
11897-4
Author(s): ShengBiao Wang, Bolun Chen, Huasong Chen, Shixuan Gao, Junhao Wang, Qiansheng Feng, Tianlong Ma, Huaiyin Institute of Technology (China)
On demand | Presented Live 10 October 2021
Show Abstract + Hide Abstract
Image deblurring and inpainting are traditional image processing problems and the effects achieved for high-resolution images are not satisfactory. In recent years, convolutional sparse coding (CSC) has been received more attentions and introduced into image processing, such as blind deblurring. However, none of works address the issue containing both blur and inpainting. In this work, we propose a novel framework of CSC for simultaneous image deblurring and inpainting. First we learn a dictionary instead of applying a given dictionary for better image representation. Second we use the learned dictionary with L1 norm to regularize image. In addition, we apply an anisotropic total variation to enhance the edges of the image. Usually we use alternating direction method of multipliers (ADMM) formulation in the Fourier domain for dictionary. We demonstrate the proposed training scheme for simultaneous image deblurring and inpainting, achieving state-of-the-art results.
11897-5
Author(s): Yuqi Han, Tsinghua Univ. (China)
On demand | Presented Live 10 October 2021
Show Abstract + Hide Abstract
The monocular camera is widely used in robots and unmanned vehicles system because it is low cost and easy to calibrate and identify. However, the depth lack of the monocular camera hinders positioning and determining the real size of obstacles in the unmanned vehicle system. To solve the problem, we propose a collaborative structure to accurately acquire the position of static or dynamic obstacles based on the partially observing information from multiple monocular cameras. After that, a reinforcement learning based obstacle avoidance algorithm is proposed for unmanned vehicles under an unknown environment. Specifically, we discuss the influence of obstacles' moving orientations on the performance of obstacles adaptive avoidance. Simulation results verify the feasibility of the proposed algorithm.
11897-6
Author(s): Yan Sun, Chang Wang, Qin Xie, Jinlei Zhang, Xinyu Liu, Zhenrong Zheng, Zhejiang Univ. (China)
On demand | Presented Live 10 October 2021
Plenary Session II
In person / Livestream: 10 October 2021 • 15:30 - 16:20 China Standard Time
15:30: Welcome and Introduction

Q&A period will follow after the talk
11905-504
Author(s): Peter L. Knight, Blackett Lab., Imperial College London (United Kingdom), UK National Quantum Technology Strategic Advisory Board for UKRI (United Kingdom)
In person / Livestream: 10 October 2021 • 15:35 - 16:05 China Standard Time
Show Abstract + Hide Abstract
The UK through a mix of government and industry funding has committed more than £1Bn over 10 years to a coordinated programme in quantum technology. Seven years into this programme I will describe here how we got there, and our goals for the future. The UK National Quantum Technology Programme has induced a step change in the UK’s capabilities for pushing forward a new sector in future information technologies. I describe how the programme arose and the activities it supported and influenced to deliver these new capabilities, building on a first phase almost £480M investment across several UK government agencies. The UK programme is now in its second phase, with a further substantial investment by UK government and global industries in the UK making a total of over £1Bn. I will describe our plans for ensuring the advanced quantum science and demonstrator platforms in imaging, sensing, communications and computing developed over the past seven years will drive the formation of the QT sector and embed quantum tech in a broad range of industries. References 1. “Blackett Review: The Quantum Age: technological opportunities.” Government Office for Science, Available: https://www.gov.uk/government/publications/quantum-technologies-blackett-review; https://uknqt.ukri.org/files/strategicintent2020/; Peter Knight and Ian Walmsley 2019 Quantum Sci. Technol. 4 040502
Session 2: Hyperspectral and Single-Pixel Imaging
In person: 10 October 2021 • 16:30 - 17:30 China Standard Time
Session Chair: Hui Qiao, Tsinghua Univ. (China)
11897-8
Author(s): Qin Xie, Chenning Tao, Xinyu Liu, Yan Sun, Chang Wang, Jinlei Zhang, Zhenrong Zheng, Zhejiang Univ. (China)
On demand | Presented Live 10 October 2021
Show Abstract + Hide Abstract
Spectral imaging can capture both spatial and spectral data of a scene, providing an efficient technique for analysis and identification. To improve the efficiency of data acquisition, compressive sensing (CS) methods have been introduced into spectral imaging systems. In this work, we propose a novel macropixel segmentation method to realize effective and non-mechanical single-pixel multispectral imaging. A series of macropixel-based patterns are designed to modulate data cube of target object. Spatial light modulator (SLM) and multispectral filter array are utilized to generate such patterns. CS algorithm is used to recover data cube from 1-D signal acquired by a single-pixel detector. Alignment of binary patterns with the subareas of macropixel filter array is conducted in the experimental set-up. Without mechanical or dispersive structure, the proposed method holds great potential in miniaturization and integration of spectral imaging devices.
11897-9
Author(s): Qian Zhou, Bo Zhao, Run Tan, Shanshan He, Lin Liu, Zhe Zhang, Beijing Jiaotong Univ. (China)
On demand | Presented Live 10 October 2021
Show Abstract + Hide Abstract
Perovskites have been widely used in solar cell manufacturing due to their extraordinary photoelectric characteristics. In this work, we have developed a kind of MHSI system to analyze the 3D data of perovskite mono-crystals. The experimental results illustrate that the specific absorption wavelength of perovskite is proportional to the thickness of mono-crystal. As the thickness increases, the absorption wavelength will shift red. In addition, the composition ratio of perovskite has a certain dependence on the absorption of perovskite. The higher the proportion of Br atoms is, the weaker the light absorption is. Therefore, MHSI technology can effectively monitor the preparation process and quality analysis of micro- and nanoscale materials and structure, and has a broad application prospect in the field of materials science and medical fields.
11897-10
Author(s): Run Tan, Zhe Zhang, Bo Zhao, Qian Zhou, Shanshan He, Lin Liu, Qiuhong Cui, Beijing Jiaotong Univ. (China)
On demand | Presented Live 10 October 2021
Show Abstract + Hide Abstract
Microscopic Hyperspectral Imaging(MHSI) technology has been used to the detection of fluorescent molecules due to its high spectral resolution. However, the low spatial resolution of MHSI makes it difficult to conduct high-precision molecular research. In this work, a two-step data processing method was proposed to enhance automatic classification effect of fluorescent molecules. We used a microscopic hyperspectral system to image the mixed five kinds of fluorescent molecular samples in transmission mode. Additionally, according to the difference of unit slope of spectral curve between different substances, an image segmentation method based on minimum transmission point is proposed. Finally, calculating the relative absorbance as the classification feature. This feature was verified on six kinds of classification models. In conclusion, the proposed two-step data processing method is suitable for the classification of multi-kinds of fluorescent molecules.
Session 3: Biomedical Imaging and Microscopy
In person: 11 October 2021 • 10:00 - 11:50 China Standard Time
Session Chair: Jiamin Wu, Tsinghua Univ. (China)
11897-11
Author(s): Shuwang Chen, Yun Wang, Meng Wang, Hebei Univ. of Science and Technology (China)
On demand | Presented Live 11 October 2021
Show Abstract + Hide Abstract
Medical CT image amplification and reconstruction system based on deep learning.Shu wang Chen, Yun Wang, Meng Wang .Institute of Information Science and Engineering, Hebei University of Science and Technology, Shijiazhuang. SRGAN (Super-Resolution Generative Adversarial Networks) algorithm is used in the system and the medical CT image super-resolution reconstruction is completed. The medical CT image processed by this method can better reflect the various details of the image, which is conducive to the observation and correct diagnosis for the doctors.
11897-12
Author(s): Aojie Zhao, Xianlin Song, Nanchang Univ. (China)
In person: 11 October 2021 • 10:30 - 10:50 China Standard Time
Show Abstract + Hide Abstract
We build a virtual simulation platform for compressed sensing photoacoustic tomography by combining compressed sensing reconstruction algorithms with photoacoustic imaging based on the k-wave simulation toolbox. On the one hand, compressed sensing can reduce sample rates, accelerated the speed of imaging. On the other hand, it can modify the demands for hardware devices and facilitate to transmit and store of data. The k-wave simulation toolbox is used to build simulation models for simulating the propagation of photoacoustic fields, recording of photoacoustic signals, and image reconstruction. We validated the performance of the simulation platform by imaging the vascular network. The results show that the virtual simulation platform compressed sensing photoacoustic tomography can achieve high-quality photoacoustic imaging with less data. The virtual platform can provide theoretical guidance for the application of compressed sensing in photoacoustic imaging.
11897-13
Author(s): Jinlei Zhang, Xiao Tao, Qin Xie, Yan Sun, Zhenrong Zheng, Zhejiang Univ. (China)
On demand | Presented Live 11 October 2021
11897-14
Author(s): Shuming Jiao, Jun Feng, Peng Cheng Lab. (China)
In person: 11 October 2021 • 11:10 - 11:30 China Standard Time
Show Abstract + Hide Abstract
A single-pixel imaging system is designed for performing a non-imaging task of all-optical logic gate operations for the first time. The spatial light intensity distribution in the imaging plane is first encoded based on the input binary bits and the type of logical gate operations. Then the synthesized binary image is sequentially illuminated by a set of all possible encoded patterns. The maximum value in the single-pixel intensity sequence will indicate the output binary result. Various logic gate operations such as AND, OR and XOR can be implemented individually or simultaneously. The proposed system can work under both coherent and incoherent lighting conditions with a simple and low-cost architecture.
11897-37
Author(s): Lihua Shen, Bote Qi, Rui-pin Chen, Zhejiang Sci-Tech Univ. (China)
On demand | Presented Live 11 October 2021
Show Abstract + Hide Abstract
The image reconstruction through a scattering media has attracted a lot of interest due to its potential application in corresponding fields. Recently, the deep learning techniques have been introduced into the computational imaging through scattering media and obtained good results. In this work, a modified U-Net model is designed under the framework of PyTorch, MobileNet is used as the backbone model. The network is trained by using mean square error (MSE) loss function, and the features of image can be extracted and the information of every pixel of the speckle field can be classified and restored in this model through depth separable convolution. The experimental results show that this network has good generalization ability for image reconstruction and improves the speed.
Poster Session
In person: 11 October 2021 • 13:00 - 14:30 China Standard Time
Conference attendees are invited to attend the poster session on Monday afternoon. Come view the posters, ask questions, and network with colleagues in your field. Authors of poster papers will be present to answer questions concerning their papers. Attendees are required to wear their conference registration badges to the poster session.

Poster Setup: Monday 10:00 to 13:00
View poster presentation guidelines and set-up instructions at
https://spie.org/PA/poster-presentation-instructions
11897-1
Author(s): Dongsheng Xu, Yantai Univ. (China); Weiwei Feng, Yantai Institute of Coastal Zone Research (China)
In person: 11 October 2021 • 13:00 - 14:30 China Standard Time
Show Abstract + Hide Abstract
Underwater optical imaging is of great significance to the development of marine resources and aquaculture. When light wave propagates underwater, light is absorbed by water body, and suspended particles in water can cause serious optical absorption and scattering. The traditional method is to obtain the intensity image of the target by using the light source, which has low contrast and poor imaging quality .An experimental platform is built based on the wave plate modulation method. The experiment shows that polarization imaging can effectively eliminate the influence of background scattered light, improve the imaging contrast and improve the imaging quality. At the same time, taking sea cucumber as the target, the device is compared with the existing focal plane polarization imaging camera, the results show that the wave plate modulation polarization imaging can effectively improve the image resolution.
11897-15
Author(s): Yonglong Sang, Jun Han, Shanghai Univ. (China)
In person: 11 October 2021 • 13:00 - 14:30 China Standard Time
Show Abstract + Hide Abstract
In order to improve the accuracy of semantic segmentation of transmission line defects, this paper proposes a segmentation algorithm based on an improved DeepLabV3+ network. The algorithm firstly expands the core network of Xception by expanding branches, removing some of the residual structures in it and reducing the network parameters. Secondly, atrous spatial pyramid pooling (ASPP) module is improved in a cascading manner to expand the perceptual field without sacrificing the feature space resolution. The decoding side fuses the three different scales of low-level semantic features extracted from the coding side to recover more spatial information and low-level features lost in the downsampling process. The experimental results show that the algorithm not only improves the accuracy of target segmentation and detection speed, but also improves the detail processing ability of defective targets in the detection of transmission line defect datasets.
11897-16
Author(s): Zhao Yingran, Xi'an Technological Univ. (China)
In person: 11 October 2021 • 13:00 - 14:30 China Standard Time
11897-17
Author(s): Bin Tai, Xiaojian Hao, Jia Wang, Haoliang Sun, North Univ. of China (China)
In person: 11 October 2021 • 13:00 - 14:30 China Standard Time
Show Abstract + Hide Abstract
Aiming at the high temperature and high pressure fireball produced by the explosion of thermobaric explosive and other energetic materials, in order to obtain the experimental data of various indexes of fireball and analyze the temperature of the produced flame, it is necessary to make real-time synchronous imaging of the state of the fireball when it is fully burned. In this paper, through the programming of software control program and the construction of overall hardware equipment, STM32 single chip microcomputer and OV7725 CMOS digital image sensor are used for multi-channel imaging. On this basis, the MV-CA013-20GC CMOS industrial camera of Hikvision company is used for high-resolution multi-channel imaging, and STM32 MCU is used for external trigger to make the multi-channel camera for image acquisition, and the delay time can be specified to achieve more accurate synchronous imaging requirements.
11897-18
Author(s): Haoliang Sun, North Univ. of China (China); Xiaojian Hao, North Univ of China (China); Jia Wang, Bin Tai, Yangcan Zhao, North Univ. of China (China)
On demand | Presented Live 11 October 2021
Show Abstract + Hide Abstract
In this paper, the temperature of butane flame is measured by Thin Filament Pyrometer(TFP). The luminous length of butane flame is about 80mm. The high temperature resistant material tungsten-rhenium wire is placed into the flame, the diameter of tungsten-rhenium wire is 0.1mm, 0.25mm, 0.5mm.A scientific sCMOS camera calibrated by emissivity calibration experiment was used to measure the radiance of tungsten-rhenium filament, and Planck's blackbody radiation law was used to calculate the temperature of the filament., so as to get the temperature of the butane flame. The results show that the highest temperature of the butane flame measured by the filament pyrometer is 1122K.Standard armored K-type thermocouple is used to verify the experimental accuracy, and the measurement error between the calculated value and the standard value is less than 5%. The experimental results of butane combustion show that this method can be applied to similar temperature measurement.
11897-19
Author(s): ShengBiao Wang, Nina Hua, Jian Li, Yuanyuan Zheng, Pengfei Xie, Junhao Wang, Huasong Chen, Huaiyin Institute of Technology (China)
On demand | Presented Live 11 October 2021
Show Abstract + Hide Abstract
In this paper, we address the rain streak removal from a single image. In order to efficiently detect and remove the annoying rain streaks, we propose a global single-directional gradient prior with L0 norm to model the rain streak. To preserve the abundant information of background, we learn a convolutional sparse coding to represent the background. Furthermore, we develop an alternating direction method of multipliers (ADMM) to solve multi-variable optimization problems. Experiments on synthesized and real-world images show that the proposed method outperforms state-of-art methods in terms of rain streak removal and background preservation.
11897-20
Author(s): Minjie Wan, Yunkai Xu, Qinyan Huang, Weixian Qian, Guohua Gu, Qian Chen, Nanjing Univ. of Science and Technology (China)
On demand | Presented Live 11 October 2021
Show Abstract + Hide Abstract
Infrared (IR) small target detection in a single frame is a challenging task due to the lack of texture and color information and the interference of background clutters. In light of the two-dimensional Gaussian-like shape of IR small target, two properties from the perspective of local gradient and directional curvature (LGDC) are characterized. Specifically speaking, the local gradients in four quadrants as well as the curvatures from four directions should distribute in a regular way in the target region. Therefore, an LGDC map is computed from the input IR image so that the contrast between target and background can be greatly improved. By this means, we are able to extract the IR small target by a simple threshold related to the mean and standard deviation values of the LGDC map. Experiments implemented on real IR images verify that the proposed method can achieve satisfactory performance in terms of local contrast enhancement and background clutter suppression.
11897-21
Author(s): Qian Xu, Xiaobing Chen, Shaozhang Xiao, ShengBiao Wang, Huaiyin Institute of Technology (China)
In person: 11 October 2021 • 13:00 - 14:30 China Standard Time
Show Abstract + Hide Abstract
In the process of 3D reconstruction of binocular vision for the production of the machine elements, local stereo matching is easily affected by illumination distortion, low texture area with low accuracy. This paper presents a weighted dynamic adaptive window stereo matching algorithm based on pixel gradient value. In order to accurately divide the high, medium and low texture regions on the image, the method of combining Sobel gradient information and phase information is adopted in this paper. It is proposed to convert RGB to HSV color space to set color threshold and to set dynamic matching window with regional threshold. The calculation function of HAD cost is established and the traditional Census algorithm is improved to perform cost aggregation. Experimental results show that the proposed algorithm is effective, has high matching accuracy, and has good robustness to optical distortion and edge information conditions.
11897-22
Author(s): Ma Zhen, Jia Yu, Huiping Liu, Ouyang Feng, Ocean Univ. of China (China)
On demand | Presented Live 11 October 2021
Show Abstract + Hide Abstract
underwater images are susceptible to the adverse effects of complex water conditions, distortion is one of the key issues that affect the quality of underwater imaging. it is important to acquire underwater images with accurate colors for marine ecological research. This paper proposes an underwater imaging technology based on a light-field camera array. The experimental system includes a camera array that simultaneously captures images of underwater targets with different exposure times, calculates the changes of tristimulus values, compensates for the color of the target image, and synthesizes a high dynamic range image. The results demonstrate the feasibility of our method.
11897-23
Author(s): Yi Yao Hu, Qipu Tan, Mao Ye, Xiaoxi Chen, Univ. of Electronic Science and Technology of China (China)
On demand | Presented Live 11 October 2021
Show Abstract + Hide Abstract
The traditional zoom imaging system composed of glass lens, mechanical control module and image sensor realizes zooming by mechanically adjusting the distance between each lens group. The combination of the adjustable focal length lens and the traditional lens can overcome the shortcomings of the traditional zoom system, such as complex structure, inconvenient operation, and discrete magnification.This paper proposes a new zoom imaging system that uses a combination of a liquid crystal lens array and a glass lens. In this paper, the imaging and zooming principle of the proposed system are explained and theoretically deduced. In addition, experiments were carried out based on the theory to test the magnification of the system under different voltages applied to the liquid crystal lens array. The magnification and image quality of the nine aperture regions of the liquid crystal lens array were also compared and analyzed.
11897-24
Author(s): Bo Feng, Ren Kun, Qingyang Tao, Han Honggui, Beijing Univ. of Technology (China)
On demand | Presented Live 11 October 2021
Show Abstract + Hide Abstract
Waste classification based on deep neural networks is up against the dataset deficiency. However, that is too expensive and time-consuming for collecting and labeling waste samples. We proposed an improved ResNet-18 model based on Model Agnostic Meta-Learning (MAML) to improve classification accuracy with a few-shot waste classification dataset. the feature extraction part of the improved model includes a convolution layer and four residual blocks; the classification part of the improved model includes a max-pooling layer and three fully connected layers. Moreover, GroupNorm is adopted to reduce the impact of different feature distributions normalization on the classification accuracy. With initial parameters from the MAML training on the Mini-ImageNet dataset, the model improve accuracy only with one training iteration results on few waste samples. The experiments verified the effectiveness of our model on the Mini-ImageNet dataset and a few-shot waste classification dataset
11897-25
Author(s): Jiaqi Xi, Yi Wang, Huaiyu Cai, Xiaodong Chen, Tianjin Univ. (China)
On demand | Presented Live 11 October 2021
Show Abstract + Hide Abstract
Improving the accuracy while maintaining the real-time performance of object tracking is a major challenge for computer vision field. In this paper, an improved Similarity-Perception-Siamese (SP-Siam) network tracking algorithm based on SiamFC is proposed. The algorithm introduces squeeze-and-excitation (SE) block and residual network for similarity map based on Siamese network, adaptively recalibrates the channel characteristic response of similarity map between target and search inputs by explicitly modeling the interdependence between channels. This study also verifies the network performance on OTB tracking datasets. The experimental results show that the squeeze-and-excitation block of similarity map has brought significant performance improvement to the existing Siamese network at slight additional computational cost achieved the goal of improving network performance.
11897-26
Author(s): Zhiyong Liu, Qun Hao, Yao Hu, Shaohui Zhang, Beijing Institute of Technology (China)
On demand | Presented Live 11 October 2021
Show Abstract + Hide Abstract
To realize the registration and stitching of two point clouds with relative low overlap rate, this paper proposes a method based on curvature features and direction vector threshold. In the registration step, the curvature feature of the point cloud data is utilized to achieve accurate matching, and the Kdtree nearest neighbor search method is used to improve the matching points searching speed. In order to further reduce the registration error, the wrong point pairs are eliminated with the direction vector threshold method. The OpenMP multi-threaded parallel calculation method is used in the process of calculating the direction vector to improve the efficiency and speed. Finally, the rotation matrix R and the translation vector t between two point clouds are obtained by singular value decomposition method. Experimental results show that the proposed algorithm can effectively improves the registration accuracy and time efficiency of point cloud data with low initial overlap rate.
11897-27
Author(s): Xinyu Liu, Chenning Tao, Qin Xie, Yan Sun, Jinlei Zhang, Qiangbo Zhang, Chang Wang, Zhenrong Zheng, Zhejiang Univ. (China)
On demand | Presented Live 11 October 2021
11897-28
Author(s): Xiao Yu, Jia Yu, Ma Zhen, Ouyang Feng, Bing Zheng, Ocean Univ. of China (China)
On demand | Presented Live 11 October 2021
Show Abstract + Hide Abstract
In order to solve the problem of quality degradation of underwater image due to absorption and scattering of water body, this paper proposes a method of underwater image enhancement based on the combination of computational imaging and deep learning. The method has achieved good results in removing image blur and scattering noise. It can effectively enhance the target images in turbid water, which will allow underwater image applications to have a wider range of areas.
11897-29
Author(s): Yong Chao, Yong Song, Zhidi Jiang, Liuyan Cao, Mei Yu, Gangyi Jiang, Ningbo Univ. (China)
On demand | Presented Live 11 October 2021
Show Abstract + Hide Abstract
Compared with traditional 2D images, omnidirectional images (OIs) can provide users better immersive visual experience, which makes OIs’ quality assessment more challenging. In this work, a spherical triangle mesh and multi-channel residual graph convolution network (Muti-RES-GCN) based blind OI quality assessment method is proposed with OI’s spherical triangle mesh generation and optimization, and Muti-RES-GCN based quality predictor. A new method of spherical triangle mesh generation and optimization is designed to extract features. Second, OI’s spherical image is divided into six view areas, in which the triangle mesh nodes are classified and input to quality predictor. The quality predictor includes Multi-Res-GCN and Estimator, which can model nodes and the dependency relationship between nodes and predict OIs’ visual quality. Experimental results show that the proposed method outperforms other state-of-the-art image quality metrics on two public databases.
11897-30
Author(s): Ren Kun, Beijing Univ. of Technology (China); Lihua Luo, Beijing Friendship Hospital, Capital Medical Univ. (China); Qingyang Tao, Bo Feng, Beijing Univ. of Technology (China)
In person: 11 October 2021 • 13:00 - 14:30 China Standard Time
Show Abstract + Hide Abstract
Thyroid-associated ophthalmopathy (TAO) is a troublesome orbital disease associated with thyroid disease, affecting ocular and orbital tissues and causing permanent facial disfigurement or severe sight-threatening. Adequate treatment depends on appropriate assessment of both severity and activity of TAO. The deep learning-based orbital Magnetic Resonance Imaging (MRI) recognition can effectively aid clinical staging of TAO patients with low measurement error and high robustness. However, it is faced with a critical challenge of insufficient labeled data. This paper proposes a deep model based on meta-learning to improve the classification accuracy with a few-shot orbital MRI dataset. The feature embedding with domain knowledge and mixed loss function is built based on the orbital MRI deep feature. They are employed in the training of a classifier based on a prototype network. Experiments demonstrate that this strategy is more accurate and has good generality.
11897-31
Author(s): Zhuang Ma, Liqiang Wang, Qing Yang, Zhejiang Univ. (China)
In person: 11 October 2021 • 13:00 - 14:30 China Standard Time
Show Abstract + Hide Abstract
Presented is an approach to improve the depth reconstruction results of dark and fluffy objects with iToF depth camera. The method consists of three steps. First, the raw depth image and grey image, as well as the confidence metric is obtained using the four-phase original frames. Second, the depth information of the dark-object region which is of low confidence is updated according to the depth value of the neighboring pixels with similar gray value and high confidence metric. Third, the depth value of the dark-object region is optimized using least square iteration considering the depth gradient consistency, and using the result of the second step as an initial constraint. The approach is tested using a dark color hairpiece. The result is compared with result of on-the-shelf Microsoft Kinect. The final valid reconstruction pixel ration is significantly improved and with an accuracy of 5cm at 2m distance.
11897-32
Author(s): Yinsheng Lv, Univ. of Science and Technology of China (China)
In person: 11 October 2021 • 13:00 - 14:30 China Standard Time
Show Abstract + Hide Abstract
In this paper, a two-dimensional maximum entropy multi-threshold segmentation algorithm optimized by genetic algorithm is proposed for near real time segmentation of plume regions in images taken by multispectral cameras.Compared with the single threshold segmentation method, the multi-threshold segmentation method used in this paper can accurately segment the plume region in a relatively complex environment.Compared with the classical threshold segmentation algorithm based on one-dimensional histogram, the two-dimensional maximum entropy threshold segmentation algorithm used in this paper combines the neighborhood spatial information of two-dimensional histogram to achieve better segmentation effect.
11897-33
Author(s): Zhangfang Hu, Yanling Xia, Yuan Luo, Lan Wang, Chongqing Univ. of Posts and Telecommunications (China)
On demand | Presented Live 11 October 2021
Show Abstract + Hide Abstract
The variable head pose and low-quality eye images in natural scenes can lead to low accuracy of gaze estimation. In this paper, we propose a multi-feature fusion gaze estimation model based on the attention mechanism. First, face and eye feature extractors based on the group convolution channel and spatial attention mechanism (GCCSAM) are designed to use channel and spatial information to adaptively select and enhance important features in face images and two eye images, and suppress information irrelevant to gaze estimation. Then we design two feature fusion networks to fuse the features of face, two eyes and pupil center position, thus avoiding the effects of two-eye asymmetry and inaccurate head pose estimation on gaze estimation. The average angular error of the proposed method is 4.1° on MPIIGaze and 5.2° on EyeDiap. Compared with the current mainstream methods, our method effectively improves the accuracy and robustness of gaze estimation in natural scenes.
11897-34
Author(s): Yue Wu, Lin Zhang, Siqi Guo, Feng Gao, Limin Zhang, Zhongxing Zhou, Mengyu Jia, Tianjin Univ. (China)
On demand | Presented Live 11 October 2021
Show Abstract + Hide Abstract
In-line X-ray phase contrast imaging is a promising technology for clinical diagnosis because of its great advantage in distinguishing low contrast tissues and simple structure to implement. In order to recover the phase projections from the phase contrast measurements, conventional phase retrieval methods were developed based on assumptions such as homogeneous material, weak attenuation, and thus suffered from limited feasibility. Deep learning-based methods have been proposed for phase retrieval and great success has been achieved. In this paper, a convolutional network based on generative adversarial network is proposed to retrieve the phase projections. Phase contrast measurements of the microspheres phantom were simulated and retrieved by the conventional methods and the proposed network. Results show the superiority of the proposed network in spatial resolution and noise suppression compared with the conventional method.
11897-35
Author(s): WangCai Zhao, Can Cui, Jun Ke, Beijing Institute of Technology (China); Xiaoli Long, Guangzhou Univ. (China)
On demand | Presented Live 11 October 2021
Show Abstract + Hide Abstract
We propose a multi-frame low-resolution (LR) images input super-resolution(SR) network(MFSRResNet) based on convolutional neural networks(CNNs) and residual learning. MFSRResNet trains on DIV2K800 and test on benchmark datasets for different upscaling factors ×3, ×4 and× 8. We train MFSRResNet with five-frame LR images inputs. There are different sub-pixel shifts among input LR images. Extensive experimental results demonstrate that MFSRResNet with five-frame LR images input shows significant performance improvement both in accuracy and visual quality compared to current state-of-the-art SISR methods. Our method MFSRResNet proves the advantage of multi-frame input as for SR task, especially meaningful for large upscaling factor such as ×8,×16.
11897-36
Author(s): Fu Zhao, Shuman Du, Dong Liang, Jun Liu, Shanghai Institute of Optics and Fine Mechanics (China)
On demand | Presented Live 11 October 2021
Show Abstract + Hide Abstract
In order to achieve global optimization in conventional optical imaging systems, complex optical design is required to eliminate various aberrations simultaneously. Imaging through scattering media can be achieved by the speckle autocorrelation method based on optical memory effect in a single-shot non-invasive way. By adding a scattering medium to the imaging system, multi-aberrations can be eliminated simultaneously. As an example, in a simple optical imaging system with spatially incoherent illumination, a ground glass plate is placed between the lens and the camera as a scattering medium. Finally, multi-aberrations such as spherical aberration, coma aberration and chromatic aberration are eliminated at the same time. Therefore, scattering media can be used as a tool to optimize optical imaging systems.
11897-38
Author(s): Zhihong Zhang, Jinli Suo, Qionghai Dai, Tsinghua Univ. (China)
On demand | Presented Live 11 October 2021
Show Abstract + Hide Abstract
As a novel asynchronous imaging sensor, event camera features low power consumption, low temporal latency and high dynamic range, but abundant noise. In real applications, it is essential to suppress the noise in the output event sequences before successive analysis. However, the event camera is of address-event-representation (AER), and requires developing new denoising techniques rather than conventional frame-based image denoising methods. In this paper, we propose two light-weight learning-based methods for the denoising of event-based sensor measurements, i.e., convolutional denoising auto-encoder (ConvDAE) and sequence-fragment recurrent neural network (SeqRNN), which can be easily adapted to image-based and event-based downstream tasks' solutions respectively. Experiments based on real data demonstrate the effectiveness and flexibility of the proposed methods.
11897-39
Author(s): Bo Chen, Jiaqi Wang, North China Univ. of Science and Technology (China)
In person: 11 October 2021 • 13:00 - 14:30 China Standard Time
Show Abstract + Hide Abstract
For wavefront curvature sensor, a wavefront restoration method based on depth neural network is proposed. A wavefront restoration model based on deep neural network is established, in which the spot data collected by the curvature sensor is used as the input and the Zernike mode coefficient is used as the output. Firstly, the feasibility of the model is analyzed, and then the influence of the number of training sets and Zernike model on the accuracy of the model are analyzed. The results show that, after full training, the loss function of the model on the test set can reach 0.0039, which can accurately restore the wavefront. For the same network model, the larger the number of training sets, the smaller the value of loss function after convergence, that is, the higher the accuracy of the model.
11897-40
Author(s): Anatoli Lisih, Aleksandr Anikeev, Andrey Zhdanov, Dmitry Zhdanov, ITMO Univ. (Russian Federation)
On demand
Show Abstract + Hide Abstract
One of the conditions for the natural perception of mixed reality is the natural mutual illumination of objects of virtual and real worlds. The current study is devoted to the research of physically correct realistic rendering methods for video see-through mixed reality devices that works with a mixed space containing virtual geometry along with a cloud of points representation of the real-world and high-resolution RGBD image of the real-world that are generated by a 3D scanning device. To form the natural illumination of the virtual objects the global illumination restored from the cloud of points representing the real-world is used. Images acquired using the developed rendering method are presented.
11897-41
Author(s): Olivier Rukundo, Lund Univ. (Sweden)
On demand
Show Abstract + Hide Abstract
This work evaluates deep learning-based myocardial infarction (MI) quantification using Segment cardiovascular magnetic resonance (CMR) software. Segment CMR software incorporates the expectation-maximization, weighted intensity, a priori information (EWA) algorithm used to generate the infarct scar volume, infarct scar percentage, and microvascular obstruction percentage. Here, Segment CMR software segmentation algorithm is updated with semantic segmentation with U-net to achieve and evaluate fully automated or deep learning-based MI quantification. The direct observation of graphs and the number of infarcted and contoured myocardium are two options used to estimate the relationship between deep learning-based MI quantification and medical expert-based results.
11897-42
Author(s): Olivier Rukundo, Lund Univ. (Sweden)
On demand
Show Abstract + Hide Abstract
The extent to which the arbitrarily selected L2 regularization hyperparameter value affects the outcome of semantic segmentation with deep learning is demonstrated. Demonstrations rely on training U-net on small LGE-MRI datasets using the arbitrarily selected L2 regularization values. The remaining hyperparameters are to be manually adjusted or tuned only when 10% of all epochs are reached before the training validation accuracy reaches 90%. Semantic segmentation with deep learning outcomes are objectively and subjectively evaluated against the manual ground truth segmentation.
11897-45
Author(s): Vitalii Nezhelskii, Iuliia Golovchanskaia, Andrey Zhdanov, Dmitry Zhdanov, ITMO University (Russian Federation)
On demand
Show Abstract + Hide Abstract
Today one of the most important problems in modern Augmented and Mixed reality applications is the analysis of surrounding objects in order to trace variability of the observed environment, the behavior of which is usually unpredictable. The solution to this problem can be widely applied in many mobile applications as a tool to interact with surrounding objects in the environment. Augmented and Mixed realities are both extremely promising directions for further research of interaction with the real environment. In this paper, we suggest a method of creation of geometric data based on a series of points in order to reconstruct the surface of real objects. This allows us to ensure the interaction with virtual objects. The proposed method comprises three steps: point detection, points clusterization into multiple groups of points, depending on their location, and the creation of geometric data such as lines and surfaces.
Conference Chair
Qionghai Dai
Tsinghua Univ. (China)
Conference Chair
The Univ. of Tokyo (Japan)
Conference Chair
Zhejiang Univ. (China)
Program Committee
MIT Media Lab. (United States)
Program Committee
Xudong Chen
National Univ. of Singapore (Singapore)
Program Committee
Shanghai Institute of Optics and Fine Mechanics, Chinese Academy of Sciences (China)
Program Committee
Jingtao Fan
Tsinghua Univ. (China)
Program Committee
SenseTime Research (United States)
Program Committee
Yo-Sung Ho
Gwangju Institute of Science and Technology (Korea, Republic of)
Program Committee
Univ. of Wisconsin-Madison (United States)
Program Committee
Ivo Ihrke
Institut National de Recherche en Informatique et en Automatique (France)
Program Committee
Tohoku Univ. (Japan)
Program Committee
The Univ. of Southern California (United States)
Program Committee
Univ. of Toronto (Canada)
Program Committee
Wanqing Li
Univ. of Wollongong (Australia)
Program Committee
Xing Lin
Univ. of California, Los Angeles (United States)
Program Committee
Yuan Luo
National Taiwan Univ. (Taiwan, China)
Program Committee
Stanford Univ. (United States)
Program Committee
Imari Sato
National Institute of Informatics (Japan), Tokyo Institute of Technology (Japan)
Program Committee
Yoichi Sato
The Univ. of Tokyo (Japan)
Program Committee
Technion-Israel Institute of Technology (Israel)
Program Committee
Univ. College Dublin (Ireland)
Program Committee
Guangming Shi
Xidian Univ. (China)
Program Committee
Shanghai Institute of Optics and Fine Mechanics, Chinese Academy of Sciences (China)
Program Committee
Tsinghua Univ. (China)
Program Committee
Boston Univ. (United States)
Program Committee
Stanford Univ. (United States)
Program Committee
Feng Wu
Univ. of Science and Technology of China (China)
Program Committee
Univ. of Shanghai for Science and Technology (China)
Program Committee
ShanghaiTech Univ. (China)
Program Committee
Xiaolin Zhang
Shanghai Institute of Microsystem and Information Technology (China)
Additional Information

CLOSED


this conference is no longer accepting abstract submissions.

POST-DEADLINE ABSTRACT SUBMISSIONS

  • Submissions were accepted through 30-August
  • Notification of acceptance by 6-September

View Call for Papers PDF