Advanced TEM Characterization for Catalyst Nanoparticles Using Local Adaptive Threshold (LAT) Image Processing
- PDF / 282,685 Bytes
- 6 Pages / 612 x 792 pts (letter) Page_size
- 84 Downloads / 166 Views
1184-HH01-05
Advanced TEM Characterization for Catalyst Nanoparticles Using Local Adaptive Threshold (LAT) Image Processing Petra Bele1 and Ulrich Stimming1,2,3 1 Technische Universität München TUM, Department of Physics E19, James-Franck-Strasse 1, 85748 Garching, Germany 2 nanotum, Technische Universität München TUM, , James-Franck-Strasse 1, 85748 Garching, Germany 3 Zentrum für Angewandte Energieforschung Bayern ZAE, Division 1, Walther-MeissnerStrasse 1, 85748 Garching, Germany
ABSTRACT Metallic and non-metallic nanoparticles, usually supported on non-metallic substrates have attracted much interest concerning their application in the field of electrocatalysis. To characterize catalysts with respect to size, morphology, structure and composition (alloys or core-shell) of nanoparticles and their associated electrocatalytic activity, transmission electron microscopy (TEM) is the state of the art method. This investigation shows the advantages of advanced image processing using the local adaptive threshold (LAT) routine.
INTRODUCTION For a thorough structural TEM characterization of catalysts, represented by small nanoparticles on a matrix, one has to deal with obstacles due to image detection and image processing [1]; in the case of image detection with: i. variation of image contrast due to local thickness changes of the support material, ii. intensity variation of similar nanoparticles based on diffraction contrast, iii. weak signal-to noise-ratio due to the difficulty to distinguish particles in the subnanometer scale from the matrix, and iv. overlapping of different particles when imaged in projection. In order to overcome these problems, computer image processing methods offer a major advantage in the data evaluation process. However, computer-assisted analysis techniques of TEM images dealing with nanoscaled or even sub-nm particles have their own difficulties arising from the applied image processing routines itself [2-3]. Therefore, a function is needed to obtain the image segmentation, which involves the classification of each image pixel to one of the image parts, either object or background.
EXPERIMENTAL DETAILS To strive for the most objective results an advanced computerized image processing routine is introduced to evaluate particle size and size distribution. A more detailed description can be found in [4]. The key for the final determination of particle diameter is to use the so-called local
adaptive threshold (LAT) routine instead of the standard global threshold routine before particle picking. By using just a global threshold, one typically has to deal with loosing too much of the desired region or getting too many extraneous background pixels resulting in an under- or overestimation of the desired region. In addition, illumination changes across the image can occur, causing brighter and darker parts not correlated to the real objects in the image. LAT typically takes a grey-scale image as input and outputs a binary image representing the segmentation assuming that smaller sub-image regio
Data Loading...
