梯度与视觉显著度下的图像灰度重叠区域识别

Recognition of Image GrayLevel Overlapping Area under Gradient and Visual Saliency

采用目前方法识别图像中存在的灰度重叠区域时,没有构建图像的显著性图,存在识别精度低、查全率低和识别效率低的问题。提出梯度与视觉显著度下的图像灰度重叠区域识别方法,根据Gestalt前背景分离原则对图像中存在的梯度通道和颜色通道进行随机阈值化处理,获得对应的二进制布尔图,采用线性平均融合方法融合利用上述获取的二进制布尔图生成视觉注意图。通过分块区域分割技术识别显著性图中存在的空间位置信息,为图像灰度重叠区域的识别提供点云数据,在云计算模式中结合局部空间降噪方法消除噪声,定位去噪处理后显著性图空置区域中存在的特征点,提取灰度重叠区域的动态特征,建立对应的灰度直方图,最后利用云检测技术识别图像中存在的灰度重叠区域。仿真结果表明,所提方法的识别精度高、查全率高、识别效率高。

When the current methods are used to identify the gray overlapping areas in the image, the saliency map of the image is not constructed, which leads to the problems of low recognition accuracy, low recall and low recognition efficiency. According to the principle of Gestalt pre background separation, the gradient channel and color channel in the image were randomly thresholded, and the corresponding binary Boolean image was obtained. The linear average fusion method was used to fuse the binary Boolean image to generate visual attention image. The spatial location information in saliency map was identified by block region segmentation technology, providing point cloud data for the recognition of gray overlapping area of image. In the cloud computing mode, the local spatial noise reduction method was used to eliminate the noise, locate the feature points in the vacant area of saliency map after denoising,extract the dynamic features of gray overlapping area, and establish the corresponding gray histogram. Finally, cloud detection technology was used to identify the gray overlapping areas in the image. Simulation results show that the proposed method has high recognition accuracy, high recall and high recognition efficiency.