一种弱增强图像下的特征提取方法设计与仿真

Design and Simulation of a Feature Extraction Method in Weakly Enhanced Image

当前的显著图像特征提取方法无法增强原始图像目标特征显著性,导致提取精度偏低、耗时长,且抗干扰性能较弱。为此,提出基于高斯函数的显著图像特征提取方法。建立高斯函数,生成关于原始图像数据的特征分布金字塔,层层滤波金字塔内所有特征点,计算迭代后的特征参数。以上述参数为判定依据,判定尺度区间内的全部特征,分辨异常特征点及目标特征点,建立提取窗口,计算在上述窗口内目标特征点水平方向及垂直方向的分布金字塔,代入到高斯响应函数中计算最终的响应阈值,至此完成目标特征提取。仿真结果证明,所提方法对原始图像特征具有较强的识别性,且与原始数据的吻合性较强、精准度较高、耗用时间较少,抵抗异常数据、噪声或冗余特征点影响能力强。

To be honest, the traditional salient image feature extraction methods have defects, such as low extraction accuracy, long time-consuming and poor anti-interference, being caused by the lack of salient features of the original image. Therefore, a salient image feature extraction method based on the Gaussian function is proposed.The Gaussian function was established to generate a feature distribution pyramid about the original image data. All feature points in the pyramid were thoroughly filtered for calculating the feature parameters after iterations. All features in the scale interval were determined by the above parameters. The abnormal feature points and target feature points were distinguished, and the extraction window was also founded to calculate the distribution pyramid of the target feature points in the horizontal and vertical directions in the window, and the results were substituted into the Gaussian response function to calculate the final response threshold, achieving the target feature extraction. The simulation results show that this method has strong recognition, high accuracy, less time consuming and excellent anti-interference.