An Immune-Inspired Approach for Unsupervised Texture Segmentation using Wavelet Packet Transform

In this paper, it is described a new unsupervised approach based on wavelet packet transform for texture images segmentation. This transform is able to decompose an image not only from the low frequency parts, but also from the middle-high frequency parts, in which there is a certain amount of texture information. After the extraction of the features, a clustering is carried out, by using an immune-inspired algorithm called ARIA (Adaptive Radius Immune Algorithm), which is capable of preserving the density information of the data and determining how many different textures (clusters) are present in the image. The performance of our methodology is compared with other methods described in literature.