Evguenii V. Kurmyshev

A framework for texture classification using the coordinated clusters representation

A statistical approach based on the coordinated clusters representation of images is used for classification and recognition of textured images. The ability of the descriptor to capture spatial statistical features of an image is exploited. A binarization needed for image preprocessing is done using, but not restricted to, a fuzzy clustering algorithm. A normalized spectrum histogram of the coordinated cluster representation is used as a unique feature vector, and a simple minimum distance classifier is used for classification purposes. Using the size and the number of subimages for prototype generation and the size of the test images as the parameters in the learning and recognition phases, we establish the regions of reliable classification in the space of subimage parameters. The results of classification tests show the high performance of the proposed method that may have industrial application for texture classification.

Comparative experiment with colour texture classifiers using the CCR feature space

An approach to grey level texture analysis is extended to colour images. Three colour texture classifiers using the CCR feature space are proposed. Textural information is derived from luminance plane by means of coordinated clusters transform along with chrominance features treated separately. The classifiers differ, basically, in the use of RGB and YIQ colour spaces. The main objective of this work is to evaluate the performance of classifiers quantitatively by means of comparative experiment on a set of VisTex and OuTex colour images. The experimental results indicate that the new classifiers are fast and at least as efficient as other texture analysis techniques evaluated on the same set of images.

Image scale determination for optimal texture classification using coordinated clusters representation

The efficiency of texture image classification is certainly influenced by image scale when a feature space or a classification method is not scale invariant. An alternative approach to the scale-invariant techniques is proposed that first estimates an effective image scale and then uses it to adjust texture features to get the best possible texture image recognition and classification. We use the correlation distance between pixels as a measure of the scale of texture images. We study the performance of classification of texture images in the coordinated cluster representation (CCR) versus an image scale and the size of the scanning window used for the coordinated cluster transform. Given the number of classes to be classified in, we find that an optimal (up to 100%) classification efficiency in the CCR feature space is obtained by changing an image scale and/or the size of the scanning window in the coordinated cluster transform.

Algorithm for training the minimum error one-class classifier of images

We propose a training algorithm for one-class classifiers in order to minimize the classification error. The aim is to choose the optimal value of the slack parameter, which controls the selectiveness of a classifier. The one-class classifier based on the coordinated clusters representation of images is trained and then used for the classification of texture images. As the slack parameter C varies through a range of values, for each C, the misclassification rate is computed using only the training samples. The value of C that yields the minimum misclassification rate, estimated over the training set, is taken as the optimal value, C(opt). Finally, the optimized classifier is tested on the extended database of images. Experimental results demonstrate the validity of the proposed method. In our experiments, classification efficiency approaches, or is equal to, 100%, after the optimal training of the classifier.

Quantifying a similarity of classes of texture images.

To quantify the concept of similarity between classes of images three measures and algorithms of calculation are proposed. The first measure is calculated through the frequency of misclassification of subimages sampled randomly from images. The second one is calculated through the cross membership of the mass center of a class in a feature space. The third measure is defined through the membership of subimages, using the distance between each subimage and the mass center of a class in a feature space. We study these measures, classifying images in the coordinated clusters representation (CCR) feature space with the minimum distance classifier. A database of images of Rosa Porriño granite tiles, previously classified by three human experts, is used in the experiments. The calculated similarity between classes is in excellent accordance with the qualitative evaluation by the human experts.

Experimental prototype of a system to measure size and distribution of droplets in organic compound fogging on glass substrates

We propose an alternative test for the characterization of fog- ging by organic compound of interior automotive materials and develop a semiautomated experimental prototype of a system for the acquisition and digital processing of images of droplets on glass substrates. We found optimal conditions of illumination of samples of fogging for the acquisition of images, achieving the resolution of 2 m that is basically the limitation of the microscope used in the experiment. The list of pa- rameters that characterize fogging was extended; in addition to size, we measure the shape, orientation, and distribution of droplets over the sub- strate. The shape of droplets on a substrate is given by equivalent el- lipses. The following processes of the fogging characterization were au- tomated: substrate positioning and image sampling at points selected randomly or on a regular lattice; acquisition, storage, and binarization of digital images; histograms of shape features, orientation, size, and depo- sition density of drops. As an example, we report the statistical analysis of drops of organic compounds. Two methods for the shape description of drops are evaluated and compared also.