Exploiting the synergy between fractal dimension and lacunarity for improved texture recognition

Fractal dimension measures the geometrical complexity of images. Lacunarity being a measure of spatial heterogeneity can be used to differentiate between images that have similar fractal dimensions but different appearances. This paper presents a method to combine fractal dimension (FD) and lacunarity for better texture recognition. For the estimation of the fractal dimension an improved algorithm is presented. This algorithm uses new box-counting measure based on the statistical distribution of the gray levels of the “boxes”. Also for the lacunarity estimation, new and faster gliding-box method is proposed, which utilizes summed area tables and Levenberg–Marquardt method. Methods are tested using Brodatz texture database (complete set), a subset of the Oulu rotation invariant texture database (Brodatz subset), and UIUC texture database (partial). Results from the tests showed that combining fractal dimension and lacunarity can improve recognition of textures.

► We investigated if synergy can be created between fractal dimension and lacunarity.
► Novel and effective methods are proposed for both measures.
► Better texture recognition is achieved when both methods are combined.
► Also methods were robust in the presence of noise and rotation.
► Tests results on three texture databases are presented in the paper.