基于图像显著性的人工视觉图像处理策略

目的:在假体设备中的视觉信息处理模块引入适当的图像处理策略，优化低分辨率下的人工视觉信息，是解决假体植入者获取的视觉信息有限问题的一种可行方法。方法:基于全局亮度对比度特征的图像显著性检测算法（LC），结合颜色空间变换和视觉注意力仿真处理模型，提出一种面向视网膜假体人工视觉信息处理策略。通过两个标准图像测试数据集对图像处理结果进行评估，同时对流行病学调查结果选取的盲人常用物体图片进行仿真。结果:评估结果验证了在提取图像前景方面与原始LC算法对比的优越性，仿真结果验证了在人工视觉条件下应用的可行性。结论:本文提出的策略有助于视网膜假体植入者在日常生活场景中更好地完成物体识别等基本视觉任务。

Image processing strategy based on saliency detection for artificial vision

Objective To effectively solve the problem of limited visual information obtained by prosthetic implant users by introducing the appropriate image processing strategy into the visual information processing module of the prosthesis device for optimizing the low-resolution artificial vision information. Methods Based on the saliency detection algorithm (luminance contrast) of global luminance contrast features, a strategy of artificial visual information optimization for retinal prosthesis was proposed by combining color space conversion and visual attention simulation processing model. Two benchmark databases were used to evaluate the result of image processing. Meanwhile, the simulation of images of the frequently used objects for the blind people which were determined by epidemiological investigation was carried out. Results The evaluation results verified the superiority of the proposed strategy over original luminance contrast algorithm in extracting image foreground. The simulated result verified the feasibility of the proposed strategy in artificial vision. Conclusion The proposed strategy is helpful for retinal prosthetic implant users to better accomplish the basic visual tasks such as object recognition in daily life.