基于神经元颜色拮抗与动态编码的轮廓检测方法

发布时间：2018-03-20 10:55

本文选题：轮廓检测　切入点：颜色拮抗　出处：《中国生物医学工程学报》2017年05期 　论文类型：期刊论文

【摘要】：基于神经元的颜色拮抗特性及神经元群体的动态编码机制,实现对图像的轮廓检测。模拟视皮层下神经元的颜色单拮抗特性,引入单拮抗感受野的动态调节机制,以充分响应颜色边界和亮度边界;利用单细胞的树突极性分布,构建初级视皮层的双拮抗神经元网络,实现对特定方位的视觉刺激响应,有效提取目标轮廓;在神经元的群体感受野内,考虑神经元的动态突触连接,融合单细胞的脉冲频率响应,实现对纹理信息的抑制作用。以BSDS500图库的图像为实验对象,结果显示该方法在提取主体轮廓的过程中能有效抑制纹理信息,其对100幅图像最佳检测结果的P值指标均值和标准差为0.58±0.04,相对CORF和CO等其他对比方法,可提高轮廓提取的准确率。所提出方法可有效实现图像的轮廓检测,为利用颜色信息以及神经元之间的动态编码、实现更高级皮层的图像理解或者视觉认知提供新的思路。
[Abstract]:Based on the color antagonistic characteristics of neurons and the dynamic coding mechanism of neuronal population, the contour detection of images is realized, and the dynamic regulation mechanism of single antagonistic receptive field is introduced to simulate the color single antagonistic characteristic of subcortical neurons. In order to fully respond to the color boundary and luminance boundary, the dual antagonistic neural network of primary visual cortex was constructed by using the dendritic polarity distribution of single cell, and the visual stimulation response to specific position was realized, and the contour of target was extracted effectively. In the population sensing field of neurons, the dynamic synaptic connection of neurons is considered, and the pulse frequency response of single cell is fused to suppress the texture information. The results show that this method can effectively suppress the texture information in the process of extracting the main contour. The average value and standard deviation of P value of the best detection results for 100 images are 0.58 卤0.04, which are relative to other comparison methods, such as CORF and CO, etc. The proposed method can effectively realize contour detection of images and provide a new idea for image understanding or visual cognition of higher cortex by using color information and dynamic coding between neurons.
【作者单位】：杭州电子科技大学模式识别与图像处理实验室;
【基金】：国家自然科学基金(61501154)
【分类号】：R338;TP391.41

【相似文献】