针对水声图像中人造物体的探测问题，给出了一种基于分形分析的方法，由于分形模型可以较好地模拟自然物体，而与人工物体存在较大差距，所以以其为主要特征可以准确地将人造物体从自然背景中探测出来。本文讨论了分维的提取方法，根据分形特征将水声图像标记为人造目标区域和非人造目标区域，并对一定噪声干扰下该方法的应用进行了研究，给出了相应的实验结果。实验结果表明，分形特征可以实现人造目标和自然物体的分类，并具有一定的抗噪声性，适宜对水声图像中的目标进行探测和识别。

Considering the problem of detecting artifacts in underwater images, a fractal-based algorithm is proposed in this paper. Because the fractal model differs greatly from man-made objects but perfectly approximates the natural objects, so the algorithm based on fractal models could accurately distinguish the artifacts from natural backgrounds. The definitions and calculations of fractal dimension are discussed in this paper. The underwater images are registered as artifacts regions and non-artifacts regions by thresholding based on the fractal features. This fractal-based object detection is applied to some noisy underwater images and the corresponding detection results are presented in this paper. The experimental results indicated that the fractal features are fit for the classification of artifacts and natural backgrounds, and have somewhat low sensitivity to the Gaussian noise. So the fractal-based algorithm is appropriate to detecting the artifacts in underwater images that are often contaminated.