针对复杂背景及目标的分形特性差异，提取图像四个方向的灰度梯度，选择最小梯度大于阈值的区域进行平滑滤波，最后对分维参数进行分段线性拉伸。

Aiming at characteristic differences of background and target, four-direction gradient is extracted out and areas with large gradient are smoothed by a linear threshold filter.

youdao

该方法首先进行中值滤波消除图像脉冲噪声，然后计算图像局部熵进行阈值选择提取目标边缘，最后进行边缘连接分割出目标区域。

The method eliminates impulse noises by median filtering then extracts edges by Otsu's thresholding based on local entropy of image, connects discrete edges and detects objects regions.

youdao

文中提出的方法可以自动提取图像的感兴趣区域，从而摒弃了采用手工标识的方式选择显著区域，使区域的匹配目标更为明确；

The presented method can automatically extract the salience regions, rejecting the approach chosen by hand to mark a salience area, thus the extracted regions match the target well;

youdao

文中提出的方法可以自动提取图像的感兴趣区域，从而摒弃了采用手工标识的方式选择显著区域，使区域的匹配目标更为明确；

The presented method can automatically extract the salience regions, rejecting the approach chosen by hand to mark a salience area, thus the extracted regions match the target well;

youdao