基于双三帧差分的运动显著前景区域目标提取

Moving Salient Foreground Region Target Extraction Based on Double Three-Frame Difference

为了解决传统运动图像前景区域目标提取方法未进行双三帧差分,而导致的提取过程计算复杂、提取结果精准度差的问题,提出了一种基于双三帧差分的运动显著前景区域目标提取方法。在三帧差分算法的基础上,将具有连续性的两次或者多次图像差分结果进行二值化处理,同时以连续的四帧图像为例,通过并集手段最大程度上弥补传统三帧差分算法造成的轮廓断裂,根据此结果类推基于连续五帧的图像,随后使用连续的双三帧图像做差分处理,根据差分结果对图像部分区域进行填充,获取前景图像中的完整运动区域,再对目标帧图像做Canny边缘提取,获取出边界轮廓从而实现目标提取。仿真结果表明,所提方法可以有效的解决传统三帧差分算法带来的缺陷,并且具有目标信息提取完整、抗噪声性强的优点。

In order to solve the problem of complex computation and low accuracy of the extraction process caused by the traditional moving image foreground region target extraction method without double three frame differnce, this paper presented a method of extracting the object from foreground region of motion saliency based on double three-frame difference. Based on the three-frame difference algorithm, two or more consecutive image difference results were binarized. Meanwhile, consecutive four-frame images were taken as an example, and the contour breakage caused by traditional three-frame difference algorithm was compensated to the greatest extent by the means of sum aggregate. According to this result, the image based on five consecutive frames was analogized. After that, consecutive double-three frame images were adopted for differential processing. According to the difference results, some areas in image were filled, so that the complete moving area in foreground image was obtained. Finally, the Canny edge extraction was performed on the target frame, and then the boundary contour was obtained. Thus, the target extraction was achieved. Simulation results show that the proposed method can effectively solve the defects of traditional three-frame difference algorithm. Meanwhile, this method extracts complete target information with strong noise resistance performance.