传统的基于过渡区域提取的目标分割算法存在噪声敏感问题, 从而会影响到过渡区域提取的准确性。 与可见光图像相比, 红外图像特别是红外光谱图像, 受到探测器无法消除的热噪声影响, 传统的目标提取算法准确率普遍降低。 此外, 虽然通过边缘能够精确定位目标, 但是无法获取目标完整边缘。 而过渡区域的灰度分布特点是可以解决基于边缘的目标提取难题。 因此为了提高目标提取的抗噪性和准确性, 提出了一种将过渡区域提取与边缘检测结合的自适应红外目标提取方法。 首先利用像元空间邻域信息构造密度, 以此有效降低噪声影响和获取图像边缘信息。 然后基于像元密度信息最大分离目标边缘与背景, 得到有效边缘和过渡区域, 进而以此生长出目标。 将边缘与过渡区域结合, 可以很好地抑制噪声, 多幅复杂场景实验评估了该方法的抗噪性能, 结果显示, 提出的方法在噪声的干扰下能较好的提取目标。

Traditional target segmentation based on transition region is sensitive to noise, which can affect the accuracy of the extraction. Compared with the visible image, the thermal noise caused by the detector in infrared image would degrade the detection rate of traditional target extraction method. In addition, although the target can be accurately positioned through the edge, it is impossible to obtain the complete edge of the target. However, the gray distribution of the transition region can solve the problem of edge. Therefore, in order to improve its anti-noise and target extraction performance, an adaptive target extraction method of infrared image utilizing edge and transition region is proposed. First of all, the pixel’s density with inform of spatial neighbors is calculated to reduce noise and obtain edges. After that, the separation between object and background is made to get transition region with edge, which is utilized to grow up the whole object. Finally, performance of anti-noise is estimated by extracting objects in several complex infrared scenes. The results show that the proposed algorithm is effective under man-added noise conditions.