泛研网

公众版 查看说明
全球科研项目数据库2026版

人工视网膜假体视觉下图像信息优化策略及其认知机制研究

项目来源

国家自然科学基金(NSFC)

项目主持人

柴新禹

项目受资助机构

上海交通大学

项目编号

61773256

立项年度

2017

立项时间

未公开

研究期限

未知 / 未知

项目级别

国家级

受资助金额

64.00万元

学科

信息科学-人工智能-认知与神经科学启发的人工智能

学科代码

F-F06-F0609

基金类别

面上项目

关键词

视觉计算 ; 视觉信息处理 ; 视网膜假体 ; 深度知觉 ; 视觉注意 ; visual information processing ; visual computing ; retinal prothesis ; visual attention ; depth perception

参与者

陆燕玉；李恒；芦卓凡；李谦；曹晓飞；苏小帆；纪晓勇

参与机构

上海交通大学

项目标书摘要：人工视网膜假体已成为盲人视觉功能修复的有效手段。然而，由于受到当前制造技术的制约，可植入的电极数量受限，当前假体植入者仅能获得低分辨率、低灰度级、缺乏颜色信息的人工视觉，因此植入者的视觉感受中缺失了大部分的纹理、对比度及空间深度等特征信息；同时，由于受到视网膜凹形结构和手术安全的限制，电极阵列尺寸不能过大，目前只覆盖黄斑区以内的视网膜区域，导致假体植入者仅能获得小视野的假体视觉。针对这些关键问题，本课题基于视觉科学、认知科学及信息科学，拟开展三方面的研究：复杂场景下基于场景对象分级的假体视觉呈现；扩大假体视觉感知视野的图像处理策略及其空间认知机制；假体视觉下的空间深度感知及其优化编码策略。旨在通过假体视觉下的信息优化处理及认知机制研究，改善假体植入者的视觉感受，提高其完成视觉任务的能力，为视网膜假体中的图像处理编码方案提供重要的理论基础，同时，也为视觉功能修复的认知机制奠定科学的实验依据。

Application Abstract: Artificial retinal prostheses have been an effective approach for restoring the blind’s visual function.However,due to the constraints from manufacturing technology,the implantable electrode number is limited,current prosthesis recipients only obtain low-resolution,low-brightness,and color-lost prosthetic vision perception.Therefore,a large amount of the feature information such as texture,contrast,and spatial depth information has lost in the recipients’visual experience.Meanwhile,due to the limitations in concave structure of the retina and surgical safety,the size of the electrode array in a retinal prosthesis can only cover a small area of the retina(within the macular area),leading to prosthesis recipients only sighting a small visual field.For these problems,based on visual science,cognitive science and information science,this research project will carry out three studies:prosthetic vision rendering based on scenario-object-classification under complex scenes;image processing strategies for expanding the perceived visual field of prosthetic vision and its spatial cognitive mechanism;and spatial depth perception and its encoding optimization strategies under prosthetic vision.The objective is to improve prosthesis recipients’vision perception,enhancing their ability to accomplish visual tasks,and contributing to the design of information processing modules for retinal prostheses by information optimization processing and the cognitive mechanism research under prosthetic vision.At the same time,it will also provide scientific experimental basis for the cognitive mechanism of visual function restoration.

项目受资助省

上海市

项目结题报告(全文)

失明是对人类生活质量影响最严重的一种残疾。通过电刺激作为神经调控手段来刺激残存的视网膜神经元，从而实现视觉功能的修复已经成为神经科学与工程领域新的研究热点。视觉功能修复在取得重大进展的同时，也存在着视觉感受分辨率低、视野小等瓶颈问题。本课题针对上述关键科学技术问题，创新地提出了利用有限数量的微电极实现高效的功能性假体视觉。为了实现这一目标，本课题开展了一系列富有成效的研究工作，首先，针对复杂场景的视觉任务，建立了基于场景对象分级的显著性检测与分割模型，实现了场景中不同重要度图像信息的分级呈现，使得假体植入者通过有限数量的光幻视点获得更加丰富的视觉信息，提高其完成复杂视觉任务的能力；在此研究基础上，针对当前视网膜假体临床应用中的视野小问题，开展扩大假体视觉感知视野的图像处理策略研究，通过基于图像内容重定位的研究思路，尽可能保留场景中的重要信息，并压缩不重要的背景信息，使得假体植入者在有限的视野范围内能够获得更多重要的视觉信息，从而实现扩大假体视觉感知视野的目标。同时，探索扩大假体视觉感知视野下的空间认知机制，为假体视觉下大脑可塑性的研究提供科学的实验依据；此外，进一步建立了适用于假体视觉的单目深度估计神经网络模型，并在公开数据集上对模型的性能进行了定性和定量的评估，验证了模型的有效性。其研究成果不仅能够改善假体植入者的视觉感受，同时，也为视觉功能修复的认知机制奠定重要的实验依据和理论基础。

  • 排序方式：
  • 1
  • /

  • 2.A hierarchical image processing strategy for artificial retinal prostheses

    • 关键词：
    • Object recognition;Prosthetics;Ophthalmology;Complex networks;Processing;Textures;Behavioral research;Hierarchical image processing;Human visual attention;Image processing algorithm;Implantable electrodes;Limited information;Postprocessing methods;Psychophysical experiments;Retinal degenerative disease
    • Jiang, Haochen;Li, Heng;Liang, Junling;Chai, Xinyu
    • 《2020 International Conference on Artificial Intelligence and Computer Engineering, ICAICE 2020》
    • 2020年
    • October 23, 2020 - October 25, 2020
    • Beijing, China
    • 会议

    For some patients with advanced retinal degenerative diseases, retinal prostheses are used to partially restore vision currently. However, due to limited number of implantable electrodes, the visual perception produced by the implants is in low resolution and has limited information bandwidth. Computer vision is the key to overcome the limitations and to optimize the information assignment. As a result, many researchers devote to investigating image processing algorithms to optimize the visual percepts of recipients. In this study, we focus on complex scene perception and present a new approach to build a hierarchical presentation. An object detection network (Mask R-CNN) is used for foreground objects extraction and segmentation. Then, a post-processing method referring to human visual attention mechanism is applied to generate this hierarchical presentation. Psychophysical experiments prove that under simulated prosthetic vision, the proposed hierarchical image processing strategy has prominent advantages in complex scene perception comparing with direct pixelization, especially when overlapping occurs in the scene. © 2020 IEEE.

    ...

  • 3.Global salient object detection based on multiple visual features

    • 关键词：
    • Image segmentation;Object recognition;Feature extraction;Background noise;Intensity difference;Parallel processing;Public dataset;Salient object detection;State of the art;Visual feature;Visual information
    • Li, Zhengyi;Li, Heng;Chai, Xinyu
    • 《2019 4th International Conference on Intelligent Computing and Signal Processing, ICSP 2019》
    • 2019年
    • March 29, 2019 - March 31, 2019
    • Xi'an, China
    • 会议

    Salient object detection has become an important tool in the fields of computer vision and image processing. In this paper, we propose a novel global salient object detection model based on multiple visual features. Firstly, following the mechanism of parallel processing of visual information, we obtain three saliency maps based on colour, intensity difference, and spatial distribution. Secondly, introducing the concept of spatial gaze point of an image, we calculate three Gaussian weighted maps based on three saliency maps to constrain their background noise and get three weighted saliency maps. Finally, the final saliency map is produced by fusing three weighted saliency maps. We compare our method with 10 state-of-the-art saliency methods on a public dataset, and the results show that our method outperforms the other 10 methods.
    © 2019 IOP Publishing Ltd. All rights reserved.

    ...
  • 排序方式：
  • 1
  • /