摘要：目的 考虑到初级视通路中视觉信息传递和处理过程中的特点，提出了一种基于初级视觉通路计算模型的轮廓检测方法。方法 提出一种体现方向选择特性的CRF改进模型，利用多尺度特征融合策略来模拟视网膜神经节的图像目标初级轮廓响应；提出一种反映视觉信息时空尺度特征的时空编码机制，模拟神经节—外膝体通路对初级轮廓响应的去冗余处理；利用非下采样轮廓波变换和Gabor变换协同作用，模拟NCRF的侧向抑制特性。最后利用初级视皮层对整体轮廓的前馈机制，实现对轮廓局部细节信息的完整性融合。结果 以RuG40图库为实验对象，经过非极大值抑制和阈值处理，得到的二值轮廓图与基准图比较，整个数据集和单张图的最优平均 指标分别为0.49和0.56，结果表明本文方法能有效突出主体轮廓并抑制纹理背景。结论 面向图像处理应用的初级视觉通路计算模型，将为后续图像理解和分析提供一种新的思路。
Contour Detection Based on the Computation Model for Primary Visual Pathway
wuwei,zhoutao,Zhu Ya-ping,Fan Yingle(Hangzhou dianzi university)
Abstract: Objective The effectiveness of contour detection in a great number of applications has been well established and demonstrated widely. This operation is a fundamental for numerous vision tasks, such as image analysis and scene understanding. It can be used for image segmentation, object detection, and occlusion and depth reasoning. Many researches on contour detection responds not only to the foreground objects but also to the background textures. In this paper, a new method of “object-only” contour detection based on the primary visual pathway computation model is proposed, according to the characteristics of visual information transmission and processing in the primary visual pathway. It is expected to obtain the accurate boundaries of object in the foreground, while suppressing both the noise and the background textures in the image. Method This paper attempts to construct a primary vision path computation model to simulate the transmission and processing of visual information flows. First, in the retinal ganglion, a classical receptive field direction selection model that combines multi-scale features is constructed to obtain the primary contour response of the image target. Then, a spatiotemporal coding mechanism is utilized to streamline the redundant features in the primary contour response, in the visual pathway of the retinal ganglion to the LGN. Furthermore, the synergistic effect of NSCT and Gabor transform is used to simulate the processing effect of the lateral suppression characteristics of NCRF on the texture background information. Finally, the feedforward mechanism of the visual pathway to the primary visual cortex is combined with the visual features of the multi-visual pathway to obtain the contour response. Result The experimental images are derived from RuG40. To verify the effectiveness of our proposed algorithm, two methods (ISO and MCI) with the best contour detection results are used for comparison. After the non-maximum suppression and threshold processing, the binary contour map obtained is compared with the reference map. On the one hand, through qualitative analysis of the proposed algorithm, although the ISO method achieves a certain balance between the false detection rate and the missed detection rate in general, some parts have severely distorted contours, such as the image Lions. Although the MCI method has better balance between the false detection rate and the missed detection rate, the actual detection effect is also very good, but its suppression of the image background needs to be improved in the processing effect of some complex background images, such as Buffalo. However, the detection model proposed in this paper has the best detection performance under the premise of ensuring that the contour detection results are close to the manual detection results. On the other hand, by quantitative analysis of the proposed algorithm, the optimal average indicators for the entire data set and single map of the RuG40 library are 0.49 and 0.56, respectively. Among them, for the average detection result of the optimal parameters of the entire library, the detection model proposed in this paper is increased by 22.5% and 6.5% compared with the ISO model and the MCI model respectively; and for the mean value of the detection results of the optimal parameters for a single image, the proposed detection model is 19.1% and 7.7% higher than the ISO model and the MCI model, respectively. Conclusion This paper attempts to construct a primary visual pathway computation model to simulate the transmission and processing of biological visual information flows. Compared with the best methods of current contour detection by ISO and MCI, it can be concluded that the algorithm proposed in this paper can suppress texture background information to a large extent while achieving complete extraction of contour information. Overall, the algorithm proposed in this paper is closer to the biological vision mechanism than the previous method. The contour detection method constructed in this paper provides a new research idea for the subsequent contour detection method based on biological vision mechanism. Subsequent research can be based on the biological vision mechanism to explore how the contour detection method proposed in this paper has an impact on more advanced image understanding and perception tasks. At the same time, the contour detection method proposed in this paper can also be used to provide a solid foundation for subsequent higher-level visual perception technology.