多维视觉图像敏感区域智能标记方法仿真

发布时间：2018-08-29 11:08

【摘要】：多维视觉图像敏感区域的标记在图像分析中是十分重要的步骤,高效的敏感区域标记方法能够有效提高多维视觉图像处理速度。采用当前方法进行多维视觉图像敏感区域标记时,因受敏感区域图像复杂度的影响,当图像敏感区域面积较大时,需要对各个目标像素点进行邻域判断,标记效率极大下降。为解决上述问题,提出一种基于分水岭的多维视觉图像敏感区域智能标记方法。上述方法首先通过计算获得多维视觉图像不受反射光干扰的分量梯度,将其融合于不同尺度结构元下开闭重建的多维视觉梯度图像,对多维视觉梯度图利用最大类间方差算法自动获取阈值,用得到的二值标记图像对原始多维视觉梯度图进行强制标记,确定多维视觉图像敏感区域的边缘信息,引入水平集方法通过符号压力函数来指引多维视觉图像敏感区域结构曲线演化方向,实现对多维视觉图像敏感区域完全分割,完成敏感区域智能标记。实验结果表明,所提方法多维视觉图像分割精度较高,有效避免了等效标记的出现,具有很好的鲁棒性。
[Abstract]:The marking of multi-dimensional visual image sensitive area is a very important step in image analysis, and an efficient method can effectively improve the processing speed of multi-dimensional visual image. Because of the influence of the complexity of the sensitive region, when the current method is used to mark the sensitive area of multi-dimensional visual image, when the area of the sensitive region is large, the neighborhood judgment of each target pixel is needed, and the marking efficiency is greatly reduced. In order to solve the above problems, a watershed based intelligent marking method for multi-dimensional visual image sensitive region is proposed. Firstly, the component gradient of multi-dimensional visual image without the interference of reflected light is obtained by the calculation of the above method, and the multi-dimensional visual gradient image is fused into the multi-dimensional visual gradient image reconstructed under different scale structure elements. The multi-dimensional visual gradient map is automatically obtained by using the maximum inter-class variance algorithm, and the original multi-dimensional visual gradient map is labeled by the binary label image, and the edge information of the sensitive area of the multi-dimensional visual image is determined. The level set method is introduced to guide the evolution direction of the structural curve of the multi-dimensional visual image sensitive region by the symbolic pressure function, to realize the complete segmentation of the multi-dimensional visual image sensitive area, and to complete the intelligent tag of the sensitive area. The experimental results show that the proposed method has high segmentation accuracy and can effectively avoid the appearance of equivalent markers and has good robustness.
【作者单位】：电子科技大学计算机科学与工程学院;
【基金】：国家自然科学基金(61370203) 科学通信安全实验室基础技术(9140c110301110c1103)
【分类号】：TP391.41

【相似文献】