基于Haar小波变换的水下小型沉底人造目标分割方法

根据水下小型沉底人造目标在傅里叶变换处理中容易丢失原本就很少的细节特征、导致图像分割效果较差的问题,在统计分析图像目标特点的基础上,提出了用Haar小波变换对原始图像数据进行多分辨率处理以尽可能保留人造目标特征,然后在拟合指数型函数确定初始平滑区间的基础上,通过迭代的方式获得二值化阈值,根据图像中干扰物的特点提出了一种统计8邻域灰度信息的干扰抑制方法.利用前面得到的阈值进行二值化分割及一次中值滤波,就可得到较为满意的目标分割结果.

Underwater Small Sinked Man-Made Target Segmentation Method Based on Haar Wavelet Transform

In Fourier transform processing of sonar image, the fewer detailed features of the underwater small sinked man-made targets are always lost in result of poor target segmentation effect.On the base of statistically analyzing the features of this kind of image targets, a new target segmentation method was proposed.Firstly, multi-resolution processing of Haar wavelet transform was used to retain men-made target features as more as possible.Secondly, we fit an exponential function to confirm the initial smoothing interval and calculated the binarization threshold through iteration.Thirdly, according to the features of the interference around the targets, a statistical eight neighborhood grey information method was used to restrain these interferences.At last, the small man-made target could be segmented from the background by the above binarization threshold.Then after a median filtering, the processing result is satisfied.The proposed method is of high engineer value because of strong pertinence, stability and rapid calculating speed.