Fumiaki Tomita

A structural analyzer for a class of textures

A method is proposed for partitioning input scenes into regions in which selected picture properties are regarded as uniform. In this procedure, the usual statistical methods of inspecting average gray levels are not used, but the uniformity of complex picture properties such as shape, size, or arrangement of subpattems in the pictures is examined by a structural analysis. The analyzing process is controlled by a supervisor which gives the order of changing the depth of interest in each picture property, and successive partitioning of the pictures is done with the aim of finding an interpretable goal partition.

Statistical Texture Analysis

Statistical texture analysis computes local features parallelly at each point in a texture image, and derives a set of statistics from the distributions of the local features. The local feature is defined by the combination of intensities at specified positions relative to each point in the image. According to the number of points which define the local feature, statistics are classified into first-order, second-order, and higher-order statistics. Statistics give various measures of texture properties. Statistics need not be only for intensities. Statistics of such local geometrical features as edge, peak and valley, and spot or blob, give measures of specific texture properties. This chapter overviews the measures of texture properties given by statistical texture analysis.

Structural Texture Analysis

From the structural view of texture, texture is considered to be composed of texture elements; texture is defined by the texture elements which are arranged according to a placement rule. The analysis includes extraction of texture elements in the image, shape analysis of texture elements, and estimation of the placement rule of texture elements. Thus the structural texture analysis is complex in camparison with the statistical texture analysis, and it gives more detailed description of texture. This chapter presents the algorithms for structural texture analysis.

System and Evaluation

There are two approaches to texture analysis: statistical analysis and structural analysis. The difference is the level of description of textures. Statistical texture analysis gives simple scalar measurements of textures. Structural texture analysis, on the other hand, gives detailed structural descriptions of textures. Which level of analysis should be used? Comparatively, statistical analysis is suitable for fine microtextures and structural analysis for coarse macrotextures. Either approach alone cannot deal with all kinds of textures properly. In the first section of this chapter, adaptive texture analysis is presented as a total system for texture analysis, which uses either or both of statistical and structural analyses depending on nature of texture, to analyze variety of textures.

Shape from Texture

When an image is a projection of a three-dimensional (3D) scene, some 3D forms in the scene can be estimated from 2D features extracted in the image. Horn (1977) first proposed a method to estimate the surface orientation from the shading in the image. It is called shape-from-shading. When we observe the surface covered with texture even by a single eye, we perceive the inclination of the surface because the texture on the surface is apparently distorted according to the inclination. The role of texture as a basis for the recovery of surface orientation was first investigated by Gibson (1950), and the method to estimate the surface orientation from the apparent texture distortion is called shape-from-texture.