Extended Statistical Landscape Features for Texture Retrieval

Texture analysis is an important research area in computer vision and pattern recognition. This paper proposes a method of extended statistical landscape features (ESLF), based on the statistical landscape features method. Extended statistical landscape features represents an image function as a surface in a three-dimensional space, which is sliced by a variable horizontal plane. Four texture feature curves describing their topological properties are extracted from the three-dimensional surface. The proposed extended statistical landscape features uses the derived feature curves to characterize image texture. Systematic experimental comparison on the Brodatz texture set as well as the VisTex texture set shows that the retrieval performance and time efficiency of the proposed extended statistical landscape features are higher than statistical landscape features, which demonstrates that the proposed method have a very high texture description power.