Objective noise characterization

The noise in an electron microscope (EM) image is determined by the dose of electrons, the detection efficiency of the scattered electrons and the properties of the object under the microscope. At present, the amount of noise in EM images is often determined manually. Unfortunately, this method has proven to be inadequate for certain types of images and it does not suffice when the noise should be detected automatically. Another method is based on taking multiple images and obtaining for each pixel both the average brightness and its variance, but this method is not able to deal with images having different points of origin, caused by objects slightly drifting while the images are created. In this bachelor thesis, the standard method of dealing with noise is analysed, and a different method is developed using repeated linear interpolation and by analysing proper bin handling. This method especially yields positive results for images with a few but major brightness transitions. The developed method is illustrated on two image sets with different complexities. In addition, this thesis also supplies a mathematical handle for describing images of drifting objects. The noise in an electron microscope (EM) image is determined by the dose of electrons, the detection efficiency of the scattered electrons and the properties of the object under the microscope. At present, the amount of noise in EM images is often determined manually. Unfortunately, this method has proven to be inadequate for certain types of images and it does not suffice when the noise should be detected automatically. Another method is based on taking multiple images and obtaining for each pixel both the average brightness and its variance, but this method is not able to deal with images having different points of origin, caused by objects slightly drifting while the images are created. In this bachelor thesis, the standard method of dealing with noise is analysed, and a different method is developed using repeated linear interpolation and by analysing proper bin handling. This method especially yields positive results for images with a few but major brightness transitions. The developed method is illustrated on two image sets with different complexities. In addition, this thesis also supplies a mathematical handle for describing images of drifting objects