|
The texture analysis is important in many applications of image analysis for classification, detection and segmentation of objects. Texture is the local spatial pattern, which is difficult to define strictly. This doesn’t mean, however, that this feature can be ignored and neglected in research on information derived from images, because of its ambiguity. The theme of the study is the interpretation of true orthophoto for automatic detection of building objects and vegetation. These elements are characterized in the image by a variety of shape, color and texture. In the present study the authors attempt to prove the thesis that the texture can be a good indicator for separation of objects such as trees from building elements. Textural procedures can be divided into three categories: structural, statistical and filter based approaches. The paper is focused on the textural filters – the Gabor filters, strengthened by the use of Laws’ "texture energy". The texture energy represents the number of changes within the window in an image subjected to a particular filter variant. The Gabor filter is linear, used for edge detection. It is a generalization of the Fourier transform, its frequency representation is similar to the imaging of human visual system and may be useful in the process of describing and differentiating textures. The data used for study have been collected during a flight over the Finland town Espoonlathi. They were as follow: image data with a spatial resolution of 0.06 m, acquired with a digital camera Rollei, and laser data from the TopEye MK II system with a resolution of 30 points/m2.
|