基于图像纹理频谱的弱目标自动检测

弱目标所含信息的相对贫乏，使得基于目标的检测算法往往难以获得满意的效果，另一方面，弱目标的存在常常使其所在局部区域的纹理较之包含该区域的更大区域的纹理发生较大改变；而纹理单元及其函数纹理频谱则分别能有效地捕获这两个区域在微纹理及宏纹理上的差异，从而检测出弱目标。基于这一思想，文中提出了一种基于纹理频谱的弱目标自动检测新方法。实验证明，该方法与其它方法相比，具有定位精确、算法简单、抗噪能力强和运算速度较快的优点。

Automatic faint target detection based on image texture spectrum

The information of the faint target is so poor that make the effect of the detection algorithm based on target dissatisfactory. On the other hand, the existence of the faint target usually make the texture of its neighborhood change greatly. Texture unit and texture spectrum can get the difference of the microtexture and the macrotexture of this two regions effectively so as to detect the target. Based on this idea in this paper, a new algorithm in automatic faint target detection is advanced. At the end of this paper, experimental results show that our algorithm is more satisfactory in the accuracy of location, the complexity of calculation, the capacity of noise|resistance and the speed. ;