ID-9-P-3477 A Texture Based Algorithm for Analyzing Transmission Electron Micrograph Images of Nanoparticles for Size Estimation

This paper presents a method for image enhancement of nanoparticles obtained from Electron Transmission Microscopy based on the two-dimensional Hurst operator for detecting edges and characterizing texture and distribution information. To measure the size of a particle, a correct identification of its edges is required; therefore for nanoparticle image analysis the edge detector to use should be able to detect weak edges and also have good noise immunity. A texture sensitive detector may help with the correct identification of each structure and assist to separate the components. In this work, we describe a new image-processing algorithm, based on the local two-dimensional Hurst operator, to improve image quality and better define the edges of the nanoparticle for later size measurement. The local two-dimensional Hurst operator uses a two-dimensional range-based neighborhood operator based on a “local Hurst operator” to extract in one operation both the edge and texture information from an image. First, a de-noising stage is performed to the original image to improve quality. Noise reduction was achieved by using wavelet transformations. Several nanoparticle samples were prepared at our Centro; these nanoparticles were composed of zerovalent or sulfate iron and carboximethyl cellulose (CMC). Images were recorded digitally with a FEI Tecnai Spirit Twin TEM operated at 80kV. Nanoparticles images samples were processed using the new image-processing algorithm in order to determine its performance. Fig. 1 shows a sample of resulting images obtained before and after image processing. Images are shown in 1024x1024 pixel resolution obtained after scaling the resulting images for visibility. As it can be seen in Fig. 1 a good improvement in image quality has been obtained. Depending of the image, the information content has been enhanced; the original structures of the particle are now clearly visible and discernible furthermore after the final stage the edges of the particle are well defined, having now a clear separation between the background and the particle. Therefore, a more precise and reliable size measurement can be performed in a next step. Experimental results show that the method is not only robust and repeatable, but it can also accommodate both nearly spherical and more irregularly shaped nanoparticles of different sizes and configurations. The results obtained suggest superior performance to previous image processing techniques and may provide a very useful tool for nanoparticle research. The algorithm was relatively easy to implement using modern software tools (such as Matlab or LabVIEW), further developments of the system will include the automated measurement of the nanoparticle size and separation of structures of interest.