Automatic bioindicator images evaluation

Abstract

Presented paper deals with processing of electron-microscope images of Picea Abies stomas. A stoma character strongly depends on the level of air pollution in the area where the tree grows - as the stoma epidermis covers with epiticular waxes to protect itself against the negative environmental influences. According to the level of incrustation it is possible to distinguish five classes of stoma structure. An automatic algorithm recognizing a degree of stoma damage, based on the microscopic image processing can be useful during the process of environmental assessment. This work is devoted to a solution of the problem of stoma evaluation by means of texture classification. There are two principles discussed in the paper: The first principle is based on gradient methods while the second one uses a wavelet transform. Possibilities of application of mentioned attitudes were investigated and the classification criteria distinguishing the stoma character were suggested, as well. The resulting algorithm was verified on a library of four hundred real images and achieved results were compared with an expert's sensual classification. Selected methods of image preprocessing as noise reduction, brightness correction and resampling were used as well.