Application Of Fast Granulometry Estimation To Texture Analysis

Texture, as an important visual information, has been a highly active researching focus in computer vision and image processing. Texture has a wide range of applications and occupies an important position in image processing. So far, research scholars have been achieved fruitful results in texture researching. There exist a wide variety of textures analytical methods, which has no shortage of excellent researching achievement. For the practical considerations of texture application, the texture image analysis is always concentrated on the extraction of the basic characteristics, using a variety of characteristics processing algorithm.The main purpose of fast granulometry estimation algorithm is the efficient extraction of the texture features. More specifically, the objective of this project is to evaluate the relevance of simplified and multiple granulometries in texture classification. In particular, we will focus on the possible benefit of combining several granulometries with diverse parameters(e.g. with variously oriented grains) in the same classification task. The main research content is focused on assessing the superiority of granulometry parameters used in texture analysis. Fast granulometry estimation algorithm is essentially an approximate of morphological opening operation, in order to get the operating results which are more approximate to opening operation with a shorter time-consuming. The Euclidean distance map getting from distance transformation is the basis for particle shape parameters extraction by granulometry at different levels, and the morphological granulometry based on opening is the theoretical basis to extract texture features. The characteristic curves finally obtained from the estimation of dilated shape parameter is essentially an extension of the granulometry curves, which are visual display of texture image characteristics.Fast granulometry estimation algorithm extracted eigenvalues from the texture images and plotted granulometry curves as an analysis basis texture image. In the experimental analysis step, there tested a mount of texture images from four well-known texture database, and it also included some tests of various language images. Furthermore, the use of polynomial regression analysis in mathematical statistics objectively and statistically evaluated the results of the experimental data of texture libraries. Experimental results shown that fast granulometry estimation algorithm has certain advantages in terms of texture feature extraction, and the curves visualize the differences of different texture shape feature.