110 likes | 270 Views
Digital Processing Techniques for Transmission Electron Microscope Images of Combustion-generated Soot. Bing Hu and Jiangang Lu Department of Civil and Environmental Engineering University of Wisconsin – Madison. Motivation and Background.
E N D
Digital Processing Techniques for Transmission Electron Microscope Images of Combustion-generated Soot Bing Hu and Jiangang Lu Department of Civil and Environmental Engineering University of Wisconsin – Madison
Motivation and Background • Quantified characterization of flame-generated soot is critical for soot research. • TEM-based study of soot properties is a reliable approach to quantifying soot size and morphology. • Limited to the quality of TEM images, this approach may be facing challenges.
Objective • By applying extensive digital image processing techniques to TEM images of soot particles, images with high qualities in senses of machine detection as well human visual inspection can be achieved. • Developed an accurate as well as efficient computational analysis of soot size and morphology based on automatic computer detection.
Typical TEM Images of Soot • Low contrast, noise • Pseudo edges caused by electron diffraction
Approach • Enhance contrast by gray level transformation. • Reduce noise by low-pass filtering. • Eliminate pseudo bright edges by blurring filtering. • Segmentation of foreground from background by thresholding. • Compensate for imperfect thresholding by morphological processing. • Identify objects by morphology processing and segmentation. • Computational analysis based on pixel value.
Thresholding • Global Thresholding • Adaptive Local Thresholding
Object Extraction and Measurement • Identify objects through extracting connected components. • Measure maximum length. • Measure projected area.
Summary and Conclusions • An economical, accurate, and rapid image processing and analysis approach has been developed for analyzing soot morphology information from the Transmission Electron Microscope images. • The techniques involved in this study include gray level transformation, convolution filtering, histogram analysis, thresholding, edge detection, image opening, extraction of connected components, and computational pixel analysis.