Abstract: A novel fractal segmentation algorithm was introduced to deal with multi-targets in sea surf. Based on analyzing and normalizing fractal morphology model, multi-targets were regarded as a whole class to avoid interaction. Then, a fractal feature named as Scale Error performs detection. After threshold segmenting and region growing, satisfied results were attained. A plenty of simulation experiments for recorded data approve its validity and veracity.

Key words: Morphology, Fractal, Scale error, Threshold segmentation, Region growing

摘要： 提出了一种提取分形尺度误差的电视图像检测分割算法。首先针对强海杂波背景特性，将多目标统归于目标类，分析了归一化的数字图像形态维分形模型，再构建分形尺度误差作为模式分类特征，并设定类间最优阈值进行分割，获得目标区域的精确位置信息，最后选取分割结果中最大局部熵点作为种子，通过区域生长得到较为完整的目标轮廓。通过对实际海景序列图像进行了仿真实验，取得了良好的检测分割效果，并具有较强的抗噪声性能。

关键词: 形态学, 分形, 尺度误差, 阈值分割, 区域生长