Description:A SYSTEM/METHOD TO COMPUTE THE TEXTON PATTERNS USING WAVELET BASED HISTOGRAM
Field of Invention
Over the past three decades, researchers have examined the benefits of texture analysis and characterization in fields as diverse as medical imaging, remote sensing, pattern recognition, industrial inspection, texture-based image retrieval, and human visual perception. The texture of an image is defined by its primitives, both in terms of the amount and variety of primitives present and their spatial organisation or layout. A random spatial arrangement is possible, as is a dependency of one primitive on its neighbour, or a dependency of 'n' primitives at once. Images of cell cultures and tissue samples taken with a microscope from aeroplanes or satellites require image processing, pattern recognition, and multispectral scanners.
Background of the Invention
Historically, there have been two main classifications of texture analysis techniques. The first of these approaches, known as the statistical or stochastic one, views textures as purely statistical events. The statistical features of pixel intensities and locations are used to define texture generation.
Stochastic formulation of texture (US20090010500A1) states that a texture can be viewed as a sample of a two-dimensional stochastic process whose properties can be defined using statistical parameters. The assumption that the interactions between pixels are mostly local is common in MRF models and other similar frameworks."The brightness level at a point in an image is highly dependent on the brightness levels of neighbouring points unless the image is simply random noise," as found by Cross and Jain. Jain and Karu, both experts in texture, stress the significance of local grey value "patterns".
Texture primitives, often known as texels or textons, are introduced in the second category, the structural approach. A texture's vocabulary of texels and a description of their relationships must be established before it can be described. The idea is to use elementary building blocks, like graphs, to express more complicated structures. Macro textures with distinct architectures lend themselves nicely to structural texture modelling. Primitive-based models have been widely employed to describe human texture perception thanks to the ground-breaking work of Julesz and Beck et al. However, this paradigm did not include grayscale textures until very recently. Recently, S'anchez-Y'anez et al. (CN101937507A) acknowledged the issue of using two different methods. Both regular and statistical features are present in any texture. Between the two extremes of perfectly periodic and perfectly random, a wide variety of textures can be found in practise. That's why it's so challenging to use a single way to categorise textures. This is why unified models have been developed, which combine deterministic and non-deterministic aspects of a texture field into one. However, Tomita and Tsuji are sceptical, stating, "There is no unique way to analyse every texture."
Wang have presented a new method of texture analysis called Texture Spectrum (TS). However, this descriptor has a wide range of possible values (i.e., 66561), and these values are not associated with one another. In order to improve rotation invariant classification over traditional LBP, a new texture feature descriptor called Integrated Logical Compact Local Binary Pattern with OR operator on Textons (ILCLBP-T) has been proposed. This descriptor reduces the texture unit size from 15 to 0. For rotation-independent texture classification, the authors suggest a new co-occurrence matrix they term the Texton and Texture Orientation Co-occurrence Matri.
Classifying the ages of stones according to their textures is an important part of texture classification. Brick, mosaic, marble, and granite texture categorization methods are proposed in the literature. The main problem is that these techniques failed to reliably categorise the aforementioned set of stones with very similar textures. This prompted the current work, which developed unique methods for texture classification based on appearance alone, including brick, mosaic, marble, and granite. Human face age categorization is another issue with texture classification algorithms. The characteristics of a person's face are considered separately for each region. Using facial traits, images of people's faces can be sorted into one of two, three, or four distinct age categories. They used computationally intensive techniques like deformable templates and snakes for localization. They are not designed for use in real-time applications. These techniques relied solely on broad categorizations, but more exact age estimates are needed for marketing purposes. New methods for determining a person's age based on their facial texture were created in this thesis after a thorough analysis of the state of the art in age classification algorithms. Many changes in shape, edges, pattern trends, and statistical fluctuations were taken into account in the present work as examples of textural features that are affected by growth.
Summary of the Invention
One of the first steps in characterizing and classifying textures is the study of patterns on textures. Textons can be used to establish a close connection between images and their constituent attributes including shape, pattern, local distribution orientation, spatial distribution, etc. The term "textons" refers to a collection of picture features, such as blobs or emergent patterns, that share a certain quality. Significant and abundant texture and shape information can be gleaned from texture patterns. Texton, one of Julesz's suggested characteristics, is a representation of texture-relevant visual patterns. Texton detection on a wavelet-decomposed texture image is performed here for the purpose of textural classification. Possible visual characteristics formed by the numerous textons. In contrast to previous research, the suggested WHTP approach tried to categorise distinct coloured stone textures based on the frequency of occurrence of textons in a wavelet-decomposed image.
Brief Description of Drawings
Figure 1: Block diagram of (WHTP) Wavelet based Histograms on Texton Patterns
Figure 2: DWT (a) First level (b) Second level
Figure.3: Six special types of Textons (a) 2×2 grid (b) TP1 (c) TP2 (d) TP3 (e) TP4 (f) TP5 and (g) TP6

Detailed Description of the Invention
The computational benefits of the wavelet approaches include quick, local, sparse, multi-resolution analysis of signals and images from the real world. This is why Wavelet based approaches are preferred when classifying and segmenting textures. It is well acknowledged that studying texture patterns is a crucial step in characterising and classifying textures. Image characteristics like shape, pattern, local distribution orientation, spatial distribution, etc. can all be closely related to textons. An image's textons are any collection of dots or emerging patterns that have a single shared characteristic. Histogram-based texture pattern extraction has not been fully explored in the wavelet domain for classification purposes.
Image classification and recognition relies heavily on being able to characterise images with derived features, yet this is a challenging problem to solve. For many texture classification and recognition tasks in the literature, efficient and accurate classification and recognition typically necessitates the computation on full image set and with vast range of grey level values. Because of this, assessing feature parameters becomes quite difficult. In order to solve this problem, the current thesis develops a "Second Order image Compressed and Fuzzy Reduced Grey level (SICFRG)" model, which decreases the picture size and grey level range without losing any major feature information. The suggested SICFRG model divides face images into five categories; this thesis uses the properties of the "Grey Level Co-occurrence Matrix (GLCM)" to determine a subject's age. Three steps are involved in deriving the SICFRG image mode of age classification. The first step is reducing the size of the original 5x5 matrix to a 2x2 second order sub matrix while preserving all of the important attributes, primitives, and local features. In the second step, fuzzy logic is used to further narrow the grayscale range of the compressed model of the image. Third, a SICFRG model is used to derive GLCM features for the image.
Texture categorization is determining the ages of stones based on their textures. Texture categorization approaches for brick, mosaic, marble, and granite have been offered in the literature. The fundamental issue is that these techniques failed to consistently classify the previously indicated group of stones with highly similar textures. This sparked the current research, which created novel methods for texture classification based solely on appearance, such as brick, mosaic, marble, and granite. Another challenge with texture classification algorithms is human face age categorization. For each region, the traits of a person's face are analyzed independently. Images of people's faces can be categorized into two, three, or four unique age categories using facial features. For this, they used computationally demanding techniques such as deformable templates and snakes.
A novel texture feature descriptor termed Integrated Logical Compact Local Binary Pattern with OR operator on Textons (ILCLBP-T) has been developed to improve rotation invariant classification over classic LBP. This description decreases the size of the texture unit from 15 to 0. The authors propose a new co-occurrence matrix called the Texton and Texture Orientation Co-occurrence Matrix for rotation-independent texture categorization.
There are a number of crucial features shared by effective wavelet transform-based texture categorization algorithms are Similar to the linear transformation, the purpose of the wavelet transform is to decorrelate the data and Orientation-sensitive information is provided by the wavelet transform, which is crucial for texture analysis. Wavelet decomposition reduces computational complexity in three important ways. In light of the foregoing, we develop a novel technique for classifying texton patterns using wavelets; we term it Wavelet based Histogram on Texton Patterns (WHTP). Figure 1 illustrates the wavelet-based texton feature evaluation method.
These days, wavelets are used in a wide variety of fields, from astronomy and acoustics to nuclear engineering and signal and image processing to neurophysiology and music to magnetic resonance imaging and speech discrimination to optics and turbulence and earthquake prediction and radar and computer and human vision and data mining and even pure mathematics applications like solving partial differential equations. In many different scientific domains, including signal processing, image compression, image segmentation, computer graphics, and pattern recognition, the Discrete Wavelet Transform (DWT) has shown to be an invaluable tool. Reduce the breadth and height of your translation by half with DWT and get a better estimate. Improving frequency resolution by doubling the number of harmonics in Fourier series expansions is essentially comparable to this procedure.
Texton frequencies are derived from the approximation and detail subbands of DWT decomposed images, at various scales, after the DWT has been applied to a collection of texture images. Texton frequencies are applied in various permutations for classification, and the best feature vectors are selected. Calculating texton frequencies for detail sub-bands of 1-level DWT decomposed images can increase the classification success rate. Discrete Wavelet Transform (DWT) is a sophisticated tool of signal and image processing that has found widespread application in many scientific domains, including signal processing, image compression, image segmentation, computer graphics, and pattern recognition.
As can be seen in Figure 2 (a), DWT is used to breakdown the image, splitting it up into four subbands that can then be sub-sampled. Fine-scale wavelet coefficients (detail images) are represented by the LH1, HL1, and HH1 subbands, while coarse-level coefficients (approximation images) are represented by the LL1 subband. Decomposing and critically sampling just the LL1 sub-band yields the next coarser level of wavelet coefficients. As may be seen in Figure 2(b), this leads to a wavelet decomposition with two levels. In a similar vein, we will employ LL2 to attain additional decomposition. This procedure is repeated until the desired magnitude is achieved. Important features demonstrated here for texture analysis and discrimination are the values in approximation and detail pictures (sub-band images). Decomposition is performed using Haar wavelets, Daubechies wavelets, and a Symlet wavelet in this case.
Textons are a type of texture primitive that must be placed according to specific guidelines. Possible visual characteristics formed by the numerous textons. The current study concludes that taking into account all possible textons is essential for achieving a precise and accurate texture categorization. There are a number of problems associated with texton size tone differential in pixel sizes among neighbours. types of textons textons that grow in only one directiontextons that grow longer but twist around. This can lead to a more subtle or pronounced form, a decrease in pre-attentional discernment, or an increase in texton gradients at texture boundaries. The six texton types used in the current thesis on a 22 grid to address this problem are depicted in Figure 3(a). Figure 3(a) shows a 22 grid with V1, V2, V3, and V4 representing the four individual pixels. In a subband image, the grid will take the shape of a texton if two adjacent pixels are highlighted in grey with the same value. Figures 3(b)–(g) depict the six distinct texton types, hereafter referred to as Texton patterns (TP1, TP2, TP3, TP4, TP5, and TP6). Haar, transform is used to first breakdown the original image. Textons are located on the approximate subband picture. Next, we look at how often each of the six possible texton orientations appears. The present study considered the total frequency of appearances of all six textons for a precise and accurate texture classification.
The proposed method WHTP conducted the tests on two Datasets. Brick, granite, marble, and mosaic stone textures with a 256x256 resolution are included in Dataset-1. These textures were gathered from the Brodatz textures, Vistex, Mayang database, and diverse natural resources captured on digital cameras. Textures of brick, granite, marble, and mosaic stone at a resolution of 256x256 are included in Dataset-2. These were gathered from Outtex, the Paulbourke colour textures database, and diverse natural materials captured on digital cameras. Dataset 1 has 80 original colour texture images, and Dataset 2 has 96.
4 Claims & 3 Figures , Claims:The scope of the invention is defined by the following claims:

Claim:
A System/method to compute the Texton patterns using Wavelet based histogram comprising the steps of:
a) Designed a technique that decompose the original image.
b) Adopted a method for identifying the textons on the image.
c) Designed a method to calculate different frequencies of different textons.
2. The Design of a System/method to compute the Texton patterns using Wavelet based histogram as claimed in claim1, anapproachHaar, transform is used
3. The Design of a System/method to compute the Texton patterns using Wavelet based histogram as claimed in claim1, a 1-level and 2-level discrete Wavelet transform is computed.
4. The Design a System/method to compute the Texton patterns using Wavelet based histogramas claimed in claim1, Adopted a method Wavelet based histogram on Texon patterns is used for precise classification of textures.