基于脊波的红外图像自适应增强算法

Adaptive enhancement algorithm for infrared image based on ridgelet

图像增强是目标检测和识别中的基本问题,为了在红外图像中能够更好地检测出目标,根据多尺度几何分析工具脊波特别适合表示各向异性和奇异性的特性,结合红外成像的特点,利用脊波对图像进行增强,并利用统计学原理对脊波系数变换原则进行自适应改进,利用统计学原理构造了目标细节特征量和脊波子带系数特征量,将这些特征量进行匹配,并且评估匹配效果,从而找出子带系数中与目标特征关联度高的部分,有针对性地对该部分系数进行放大,其余部分系数进行抑制处理,处理后的系数进行脊波逆变换还原的图像。检测结果与对比度增强、线性滤波等方法进行比较说明:该方法效果良好。

Image enhancement is the basis for target detection and identification. In order to detect infrared target well, we propose an improved algorithm by using ridgelet transform which is a kind of multi-scale geometric analysis tool and is especially suitable for describing the 2-D signals. Firstly coefficient transform principle of ridgelet transform is introduced. Then adaptive enhancement algorithm is proposed. Tensile scale and compressing scale of coefficient in ridgelet transform domain are the key points of the method. Statistics is used in processing of transform to format characteristic data for consistent match, and the algorithm is assessed with sign-test method. The algorithm improves the target＇s detail characteristics. Finally images are reconstructed and compared with those processed with other traditional methods. Our enhancement method is proven effective by experimental results.