对图像进行面向自动目标识别(Auto Target Recognition,ATR)的压缩其关键是快速而准确地检测到目标感兴趣区域ROI(Region-of-interest),并将其与背景区域分别进行不同比特率的压缩。本文将形态Haar小波法与数学形态学方法相结合来实现目标ROI的检测,设计了新的目标ROI检测算子。对采集图像进行二维形态Haar小波分解,结合目标ROI检测要求的特点,仅在尺度信号域内应用设计的目标ROI检测算子,最终完成目标ROI的检测。仿真实验表明,该方法对目标ROI的检测率最高可达到1.000 0,而最低虚警率仅为0.001 2;对含像素级别为102×102的图像,所需运算时间仅为10-1 s。与传统方法相比,本文算法对目标ROI检测效果好,运算简单,节省了运算时间和硬件资源。

The Auto Target Recognition (ATR) has been used to solve the problem that the huge data flows provided by a high speed image acquistion system are not easily transferred and stored,in which the key is how to find the Region-of-interest (ROI) of a target quickly and exactly.To detect the ROI of the target,a morphology Haar wavelet method and a mathematic morphology are combined to use in the ROI detection and a new target ROI detection operator is designed. An image is decomposed with morphology Haar wavelet,then the new ROI detection operator is used in the field of scale signal decomposed by morphology Haar wavelet to find the ROI of target.The simulation results indicate that the highest detection ratio of the ROI can reach 1.000 0 and the lowest false alarm ratio of the ROI only is 0.001 2.Moreover,the time consumption is only 10-1s for a image with a pixel level of 102×102.In comparison with traditional algorithmns,this method can find the ROI of the target effectively and can save the time consumption and hardware resource.