面向视觉假体的复杂图像处理技术

为了更准确地提取视觉假体采集的复杂图像中的信息，提出了一种基于视觉显著性的复杂图像视觉焦点检测和基于多尺度信息融合的图像边缘特征提取算法。算法首先利用GB算法计算复杂图像显著性图，然后利用双阈值分割和形态学方法提取出复杂图像中的视觉焦点区域。其次，针对传统边缘检测中单尺度微分算子的不足，提出了融合多尺度Sobel的边缘检测方法。提出的复杂图像处理策略既简化了图像信息编码的复杂度，又提高了边缘提取算法的噪声稳健性。

Complex image processing techniques for visual prosthesis

In order to extract the information in the complex image of the visual prosthesis more accurate, a new focus sub image extraction method based the visual saliency and new edge detection algorithm fusing multi scale information of image edge strength are proposed. Firstly, it uses the GB method to calculate the image saliency map, and applying the double thresholding on the saliency map, the initial segmentation is modified by the morphology processing and the focus sub image is obtained finally. Secondly, considering the deficiencies of traditional single scale edge detection, the edge detection based on the fusion information of the muti scale Sobel operators is proposed. The proposed complex image processing strategy not only simplifies the complexity of the encoding image information, but also improves the noise robustness of edge extraction algorithm.