Wavelets for texture analysis

we will outline the scheme they have proposed. In [52] the wavelet transform was used both to analyse the image prior to segmentation, enabling feature selection, as well as to provide spatial frequency-based descriptors as features for segmenting textures (see Fig. 27). Smooth and textured images can easily be distinguished from each other by examining their wavelet transforms. 

A three-level wavelet decomposition of a 2-D image results in 10 WT channels (Fig. 29). The channels are numbered from 1 to 10. The energy of each channel is calculated by finding the mean magnitude of the wavelet coefficients in this channel. The use of energy is sound and has an obvious physical interpretation. Moreover, it is additive and its total is conserved by 

The scale increa.ses from top to bottom. The second and third columns. show fF2j/(jT, v) y 4 and y) y 4. Black, grey, and white pixels stand for negative, 

In order to successfully discriminate between smooth and textured images using the energy of the different channels of their wavelet transforms. Porter and Canagarajah [52] have further grouped the ten channels into low 

In the last formula w j is a wavelet coefficient within the channel c. If the ratio p is above a certain threshold, i.e. p t, the pixel (or block of pixels) is labelled as smooth otherwise it is labelled as textured. 

---

S. Livens, P. Scheunders, G. van de Wouwer and D. van Dyck, Wavelets for Texture Analysis, Technical Report, VisieLab, Department of Physics, University of Antwerp, 1997. 

---