Texture Algorithms Research Articles

Sand grain analysis—image processing, textural algorithms and neural nets

A quantitative method for characterising and classifying quartz grain form by mathematical analysis of surface texture is presented. Scanning electron images of quartz grains were ‘frame grabbed’ and converted to a digitised grey level image. Image enhancement, segmentation and histogram equalisation were applied to produce standardised images. Two major textural approaches were then applied. The Roberts gradient operator determined the degree of surface edgeness while calculation of spatial grey level dependence matrices allowed production of distribution maps of surface homogeneity, entropy and correlation. Textural parameters were obtained from samples of 0.5 mm quartz grains from three distinct populations: Desert quartz; Fire Island, New York, beach grains; and Brazilian crushed quartz. A comparative analysis using discriminant analysis and neural networks was undertaken to quantify the success in classifying the different populations. Both methods achieved excellent degree of quartz grain classification. However the use of neural networks provided a more robust method of analysis particularly when presented with incomplete data sets.

Analysis of human impact on a forested landscape of central Italy with a simplified NDVI texture descriptor

Abstract Vegetation tends to reach a maximum of organic matter, species and complexity compatible with available resources. Consequently, forested environments with high levels of organization are usually characterized by high cover density values and a high degree of spatially homogeneous biomass at the landscape scale. There are however natural or more often anthropic disturbances which tend to oppose this trend. Because forest structure is a spatial phenomenon, quantitative measures of spatial heterogeneity are a useful means of increasing our understanding of landscape structure and dynamics. The primary objective of this study, which uses remotely sensed inputs of plant biomass, was to describe the processes that generate heterogeneity within a forested mountainous landscape of Central Italy at the scale of Landsat TM data with a simplified texture algorithm. Results show that the anthropic impact tends to oppose the natural tendency of vegetation towards a high degree of spatially homogeneous biomas...

Textural classification for ecological research using ATM images

Abstract Investigation of spectral and textural classification of high resolution ATM image of a semi-natural scene is presented. Pure spectral classification using bands 5, 7, 9 and the maximum likelihood classifier yielded 56, 63 and 64 per cent overall classification accuracies with 1-25m, 2-5m, and 50m spatial resolution data respectively. Application of combined spectral and textural classification using bands 5, 7, 9 and various texture features from seven texture algorithms ( spatial grey level dependence matrices-SGLDM, grey level run length matrices-GLRLM, busyness, neighbouring grey level dependence matrices-NGLDM, sum and difference histograms-SADH, and fractal analysis), yielded overall classification accuracies from 58-65 per cent at 1-25 m resolution. It is concluded that texturally-based classifications improve overall classification although improvements are not dramatic. The first-order texture measures from algorithms like GLDH and SADH have shown more promise than second-order algorithms, like SGLDM and NGLDM. The energy feature from most of the texture algorithms shows considerable classification potential. A selection of distance metric corresponding to the size of the spatial unit for a given cover type improves the classification of that class. With degradation of spatial resolution the overall accuracy of textural classification improves up to 69 per cent for 5-0 m resolution data.

Quantitative Texture Analysis in Two‐Dimensional Echocardiography

BACKGROUND: Myocardial reflectivity is abnormally increased in patients with thalassemia major under transfusion treatment, probably due to myocardial iron deposits and/or secondary structural changes. Such increased reflectivity has been detected by both qualitative and subjective analysis of two-dimensional echocardiographic (2-D echo) images and quantitative assessment of integrated backscatter amplitude with noncommercially available ultrasound prototypes. The purpose of this study was to assess the acoustic properties of myocardium in patients with beta-thalassemia major and iron overload by means of quantitative computerized offline textural analysis of conventionally recorded 2-D echo images, and to compare textural data with other qualitative (visual assessment) and quantitative (ultrasound backscatter analysis) approaches for myocardial ultrasound tissue characterization simultaneously applied to these patients. METHODS AND RESULTS: Thirty-five young patients with thalassemia major, without clinical signs of cardiac failure, and 20 age and sex matched normal controls were studied by echocardiography. Each patient was receiving blood transfusion every 2-3 weeks. Two-dimensional echo images, obtained with a commercially available echocardiograph using the parasternal long-axis view, were digitized off line and analyzed by first and second order texture algorithms applied to regions of interest in the myocardium (septal and posterior wall). The mean gray level value was higher in thalassemic patients than in controls on both the septum (110 +/- 25 vs 57 +/- 13, arbitrary units on a 0-255 scale; P < 0.01) and posterior wall (91 +/- 25 vs 67 +/- 18; P < 0.01). Among second order statistical parameters, contrast and angular second moment significantly (P < 0.01) differentiated septal and posterior walls of patients and controls. In thalassemic patients, no consistent correlation was found between wall texture parameters and hematologic (years of transfusions and chelation, number of transfusions), 2-D echo (posterior wall thickness, left ventricular end-diastolic diameter), and Doppler (transmitral E/A waves ratio) parameters. Myocardial walls with visually assessed increased echo reflectivity showed a trend toward higher values of mean gray level when compared with myocardial segments with qualitatively assessed normal reflectivity (septum: 121 +/- 26 vs 106 +/- 24; posterior wall: 105 +/- 23 vs 87 +/- 23). Although radiofrequency integrated backscatter has been demonstrated to be capable of identifying thalassemic patients, no significant correlation was found between mean gray level (by texture analysis) and radiofrequency data (septum: r = 0.03; posterior wall: r = 0.09; P = NS for both). CONCLUSIONS: Myocardial walls affected by hemochromatosis show ultrasound image texture alterations that may be quantified with digital image analysis techniques and appear mostly unrelated to hematologic and conventional, as well as radiofrequency-based, echocardiographic parameters. These changes in quantitatively evaluated echo reflectivity are present even before the development of clinical and echocardiographic signs of cardiac dysfunction. (ECHOCARDIOGRAPHY, Volume 13, January 1996)

Automated object recognition of blue-green algae for measuring water quality—A preliminary study

In this paper a computer algorithm for the automated detection of blue-green algae is presented. Samples of seven species of blue-green algae and two species of green algae were examined under a microscope and transferred to a computer. The microscope pictures were stored as digital images. In order to locate the organisms Image Segmentation routines were applied. Image Enhancement improved the quality and appearance of the species in the images. With the aid of shape algorithms (Fourier Descriptors, Moment Invariants) and textural algorithms (Cell finding algorithm, Statistical measures) certain features were identified, extracted and then fed into a classifier. The classifier, using Discriminant Analysis, was then able to predict 155 out of 158 samples correctly.

Exact convergence of a parallel textured algorithm for data network optimal routing problems

In our earlier paper (1991), a textured decomposition based algorithm is developed to solve the optimal routing problem in data networks; a few examples were used to illustrate the speedup advantage and the convergence conditions for the textured algorithm to converge to a global minimum. The speedup advantage is investigated in Huang et al. (1993). However, the theoretical foundation is not provided. In this paper, we provide the foundation. First, we show that for any textured decomposition, the algorithm always converges to a stationary point, which may not be a global minimum. And then, we prove that if the conditions of the exact convergence theorem are satisfied, the textured algorithm will converge to a global minimum.

A parallel HAD-textured algorithm for constrained economic dispatch control problems

In our earlier papers we developed a parallel textured algorithm to solve the constrained economic dispatch control (CEDC) problems. The exact convergence theorem and its proof were provided to guarantee the convergence of the algorithm to the true solution; and some examples were given to show the impact of exact convergence conditions. In this paper, we incorporate the hierarchical aggregation-disaggregation (HAD) concept and the textured concept to solve the CEDC. The algorithm is then implemented on an nCUBE2 machine and tested on a modified IEEE 14-bus system, a modified IEEE 57-bus system, a 114-bus system, and a 228-bus system. Some test results are given to show the speedup advantage of the proposed algorithm over the one without textured decomposition. Even when the proposed algorithm is executed sequentially, the speedup is still essential.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Investigation of Image Resampling Effects upon The Textural Information Content of a High Spatial Resolution Remotely Sensed Image

Abstract An empirical investigation into the effects of bilinear and cubic convolution resampling upon the textural information content of a high spatial resolution remotely sensed image is described. Textural feature values calculated with the grey level difference histogram algorithm are shown to be modified in a complex manner by the resampling process. The modification is found to be dependent upon the resampling technique, the parameters or the texture algorithm and upon the grey level structure of the image. These findings indicate that texlurally-based image analysis procedures may be most appropriately applied prior to image resampling.

On-line processing of transmission electron microscopic images

Some methods for an on-line image processing of transmission electron microscopic (TEM) images are described. The TV-scanning system is characterized by the intensity and modulation transfer function. Nonlinear noise filtering and iterative phase alignment for preprocessing the images are presented. Biomedical applications include a texture algorithm based on visual contour perception and a three dimensional reconstruction of serial sections.

Mammographic Texture Analysis: An Evaluation Of Risk For Developing Breast Cancer

This paper deals with early and accurate breast cancer risk assessment for women. The use of texture analysis tools for the eventual development of an automatic system is proposed. In a first step, a standard procedure for obtaining x-ray mammograms is set up, the resulting radiographic images then being classified into four risk groups by a specialist. In a second step, specific and selected texture algorithms using both global and local statistical properties of the images are implemented. A number of x-ray mammograms have been studied. One of the resulting important observations is that it seems inappropriate to define a set of distinct classes of risk; rather, an increasing gravity degree correlated to a continuous evolution of the mammographic textures from the lowest to the highest degree of risk is to be preferred. Finally, a systematic comparison between the human classification and the numerical coefficients provided by the texture analysis is performed. The coefficients do not allow risk classification by themselves. A critical examination of these preliminary results leads us to a constructive discussion concerning the future developments of the proposed method.

Three‐dimensional reconstruction of torsos by computer and its applications

AbstractComputer graphics has been applied in the past to the CAD (computer‐aided design) in machine design. However, the technique is recently applied to other fields of industry, and positive attitudes toward CAD are observed in the design of dresses. This paper proposes an algorithm to determine the three‐dimensional structure of torsos from basic dress patterns.A method to reconstruct three‐dimensional objects from dress patterns is described first, and then the algorithm to texture the torsos is described. By the texturing algorithm, it is made possible to choose cloths and design according to individual taste in the form of the completed dress worn on the torso of the customer.

A Theoretical Comparison of Texture Algorithms

An evaluation of the ability of four texture analysis algorithms to perform automatic texture discrimination will be described. The algorithms which will be examined are the spatial gray level dependence method (SGLDM), the gray level run length method (GLRLM), the gray level difference method (GLDM), and the power spectral method (PSM). The evaluation procedure employed does not depend on the set of features used with each algorithm or the pattern recognition scheme. Rather, what is examined is the amount of texturecontext information contained in the spatial gray level dependence matrices, the gray level run length matrices, the gray level difference density functions, and the power spectrum. The comparison will be performed in two steps. First, only Markov generated textures will be considered. The Markov textures employed are similar to the ones used by perceptual psychologist B. Julesz in his investigations of human texture perception. These Markov textures provide a convenient mechanism for generating certain example texture pairs which are important in the analysis process. In the second part of the analysis the results obtained by considering only Markov textures will be extended to all textures which can be represented by translation stationary random fields of order two. This generalization clearly includes a much broader class of textures than Markovian ones. The results obtained indicate that the SGLDM is the most powerful algorithm of the four considered, and that the GLDM is more powerful than the PSM.

Toward a structural textural analyzer based on statistical methods

This paper reports investigations aimed at developing a feature set for the Spatial Gray Level Dependence Method (SGLDM) which measures visually perceivable qualities of textures. In particular it will be shown that the inertia measure commonly used with the SGLDM can be used to characterize the placement rules and the unit pattern of periodic textures. In this way one may formulate a structural approach to texture analysis based on the statistical SGLDM. To mathematically verify that the features used with the SGLDM can be used to characterize the unit pattern and placement rules of a periodic texture, a mathematical tiling theory model is proposed. This model allows one to develop the mathematical machinery necessary to prove the result. In a companion paper other features which measure visually perceivable qualities of patterns will be developed. The reason for concentrating on the SGLDM for developing such a feature set is predicated on perceptual psychology experiments and comparison studies of various texture algorithms. All of these studies indicate that second-order probabilities of the type measured by the spatial gray level dependence matrices are important in human texture discrimination and that these matrices contain more important texture-context information than the intermediate matrices of other statistical texture analysis algorithms.

Texture Measurements for the Automatic Classification of Imagery

The stated purpose is to demonstrate the applicability of texture measurements for making distinctions between classes of imagery. Multispectral images obtained from aircraft and satellites have been successfully delineated into land use classes on the basis of density in the different spectral bands. However, spatial patterns can add additional information to improve classification accuracy. A comparison is made between the results obtained using five texture algorithms for separating land use classes using ERTS imagery. The transforms evaluated are the Karhunen-Loeve, the fast Fourier, the Walsh-Hadamard, the Slant, and a digital matched filter.