Nelleri Anith, Joseph Joby, Singh Kehar
Photonics Group, Department of Physics, Indian Institute of Technology, Delhi, New Delhi, India.
Appl Opt. 2006 Jun 10;45(17):4046-53. doi: 10.1364/ao.45.004046.
We demonstrate the validity of wavelet-based processing for recognition and classification of three-dimensional phase objects. One Fresnel digital hologram of each of the three-dimensional (3-D) phase objects to be classified is recorded. The electronic holograms are processed digitally to permit 3-D object information to be retrieved as two-dimensional digital complex images. We use a Mexican-hat wavelet- matched filter (WMF) to enhance the correlation peak and discriminate between the objects. The WMF performs a wavelet transform (WT) to enhance the significant features of the images and the correlation of the WT coefficients thus obtained. We compare the feasibility of a WMF-based object classifier with the matched-filter-based classifier to classify our four 3-D phase objects in a 3-D scene into true or false classes with minimal error.
我们展示了基于小波的处理方法在三维相位物体识别和分类方面的有效性。记录了每个待分类的三维（3-D）相位物体的一幅菲涅耳数字全息图。对电子全息图进行数字处理，以便将三维物体信息作为二维数字复图像检索出来。我们使用墨西哥帽小波匹配滤波器（WMF）来增强相关峰并区分不同物体。WMF执行小波变换（WT）以增强图像的显著特征以及由此获得的WT系数的相关性。我们将基于WMF的物体分类器与基于匹配滤波器的分类器的可行性进行比较，以便在三维场景中将我们的四个三维相位物体分类为真类或假类，且误差最小。
