结合Contourlet与ACPSO的红外热波图像增强

针对无损检测红外热波图像对比度低、边缘模糊、含大量噪声的问题,提出了基于Contourlet变换和混沌变异双粒子群优化(adaptive chaotic variation particle swarm optimization,ACPSO)的自适应增强方法。红外热波图像经Contourlet变换分解成低通和带通方向子带。低通子带系数依据一种适应于人类视觉系统的灰度级变换调整,待定参数由ACPSO确定,为了得到最佳增强效果,适应度函数由一种对比度测量函数确定;带通方向子带系数的调整则采用非线性增益函数实现,从而抑制噪声并增强细节。大量红外热波图像增强实验结果表明,与现有的4种增强方法相比,能大大提高缺陷和背景之间的对比度,增强缺陷的边缘细节。进一步采用倒数熵多阈值分割方法时,能更有效地提取缺陷,为后续准确进行缺陷识别和尺寸测量奠定了基础。

Enhancement of infrared thermal wave images based on contourlet and adaptive chaotic variation particle swarm optimization

The infrared thermal wave image in the nondestructive testing have the disadvantages of low contrast, blurred edges and strong noise. Thus an adaptive enhancement method of the infrared thermal wave image based on contourlet transform and adaptive chaotic variation particle swarm optimization (ACPSO) is proposed. An infrared thermal wave image is decomposed into a low-pass subband and band-pass directional subbands through the contourlet transform. Then the coefficients of low pass subband are adjusted according to a gray scale transform, which is adapted to the human visual system. The related parameters are determined by ACPSO. In order to obtain the best enhancement effect, the fitness function can measure the contrast of images. While the coefficients of band pass directional subbands are adjusted by a nonlinear gain function. Thus noise is suppressed and details are enhanced. A large number of experimental results of infrared thermal wave image enhancement show that, compared with four existing image enhancement methods, the proposed method can improve the contrast between the defects and the background greatly, enhance defect edges and suppress noise. While multi thresholding method using maximum reciprocal entropy is further adopted, the defects are exacted more efficiently. The proposed method lays the foundation for the subsequent accurate defect recognition and measurement of defect sizes.