A photoacoustic (PA) signal of an ideal optical absorbing particle is a single N-shape wave. PA signals are a combination of several individual N-shape waves. However, the N-shape wave basis leads to aliasing between adjacent micro-structures, which deteriorates the quality of final PA images. In this paper, we propose an image optimization method by processing raw PA signals with deconvolution and empirical mode decomposition (EMD). During the deconvolution procedure, the raw PA signals are de-convolved with a system dependent deconvolution kernel, which is measured in advance. EMD is subsequently adopted to further process the PA signals adaptively with two restrictive conditions: positive polarity and spectrum consistency. With this method, signal aliasing is alleviated, and the micro-structures and detail information, previously buried in the reconstructing images, can now be revealed. To validate our proposed method, numerical simulations and phantom studies are implemented, and reconstructed images are used for illustration.
A training image optimization method in multiple-point geostatistics and its application in geological modeling
