Saliha Aouat

Texture Segmentation and Matching Using LBP Operator and GLCM Matrix

Image processing is a dynamic research area. Recently, a lot of works have been made, efficient approaches have been developed and good results have been obtained. In this work, we propose a new texture matching and segmenting approach based on a new decomposing architecture. This method starts with one main window MW. For each iteration, the MW is reduced and all possible windows with the same size of the MW are generated. The Local Binary Pattern LBP operator, which is gray-scale invariant texture measure, and the Gray Level Co-occurrence Matrix (GLCM), which is a second order statistics measure, have been applied independently to extract the features from each generated window. Synthetic images and test images generated randomly from Brodatz album have been used in the experimentation. Good performances have been obtained and some results will be shown in the test section of this chapter.

Indexing binary images using quad-tree decomposition

The tree structure is introduced to specify block-oriented decomposition of database images. These decomposition structures offer a fundamental data model for specifying image content in large image databases.

A cognitive approach for texture analysis using neighbors-based binary patterns

The human brain receives images from the natural world and understands scenes, places and events quickly, outperforming the most advanced artificial vision system. Most of surfaces are textured in real life. Thus, In this paper, a novel texture analysis method has been proposed. The texture can be seen as a visual representation of complex patterns that lead to cognitive understanding of the environment. Our method is inspired from the Local Binary Pattern (LBP) method. The proposed Neighbor based Binary Pattern (NBP) extracts the local pattern from the texture using an analysis window. Each neighbor of the central pixel is thresholded by the next neighbor and encoded (starting from the top-left neighbor and going clockwise). Thus, the central pixel describes the relative pertinent information between its neighboring pixels. The rotation invariant version of the NBP method extracts patterns which are robust against rotation. For this, the encoding process starts always from the higher neighbor. The encoding process is applied on whole the original image in order to obtain the RINBP image. A histogram is calculated from the RINBP image to describe the texture. The size of the obtained histogram was reduced while keeping the relevant information. In the experiments, the performance of the proposed feature is evaluated on thirteen textured images from Brodatz texture album. It is shown that the RINBP method outperforms the earlier versions of the rotation-invariant LBP and the classical NBP method. This is due to its ability to extract the relative and relevant information from the local neighborhood.

A new approach for texture segmentation based on NBP method

Nowadays, image processing is an interesting research area due to the growth of the communication technologies. Matching problem, which consists of localizing one texture in an image, that contains several textures is one of the fundamental problem of image processing and pattern recognition. In this paper, a new feature extraction method and texture segmentation system are proposed. The proposed method (RINBP) is robust against rotation and improves the ability of extracting the local information. The segmentation architecture follows several steps. First, fixing a converging point α. After that, a Main analysis Window (MW) starting from α to the bottom left corner of the image is determined. Then, several possible windows are extracted and the feature extraction method is applied on each window. Finally, a similarity measure is calculated in order to decide if this window is pertinent or not. This process is stopped until the size of the MW reaches a minimum size. Each pertinent window increases the relevance of the desired texture in the output image. Finally, an image of relevance is obtained by considering the most relevant area. For the experiments, textured images generated from Brodatz album database are used. The experiments have shown the superiority of our method compared to other existing methods. The obtained results have illustrated the robustness and the efficiency of the proposed segmentation method based on the relevance of the analysis windows.

Facial shape-from-shading using features detection method

The shape reconstruction from one image is an important problem in the computer vision field. It is the so-called shape from shading problem SFS. It is known to be an ill-posed problem, because each pixel has a family cone of surface normals satisfying the image-irradiance equation. To make the shape from shading problem well-posed, several constraints were imposed on the surface of the 3D object, the model of the camera, and the light source which illuminates the 3D object. The idea of the proposed method is to use a facial features detection method to determine the singular points of the face in order to use them to generate the 3D object. It occurs in two main steps: the first step is the extraction of the facial features the eyes, the nose and the mouth. Using this information, we can extract the singular points which represent the maximum gray-scale value and apply the local fast-marching method to compute the depth of these points. The second step is the application of the global fast-marching to compute all points around the singular points. The proposed method is tested on real facial images.

Shape from Shading with and without Boundary Conditions

The shape from shading field attracts the attention of many researchers. Several methods have been proposed in several domains of computer vision. Two classes of methods are used: local and global resolution methods. Local methods deal with each pixel and its neighbors. Global methods, however, deal with all the pixels of the image at the same time. Other methods of resolution are integration methods which may be local or global. Integration methods, solve the problem of shape from shading throw two steps: the generation of the needle-map then its integration to generate the 3D object. This chapter proposes a new needle-Map integration method. The needle-Map is calculated from an image generated by a perspective camera. At first the boundary conditions was supposed to be known to solve the problem, then an improvement is performed to integrate the needle map without boundary conditions thanks to the utilization of a singular point of the image. The proposed technique was tested on synthetic and real images.

3D object indexing and recognition

In this paper, we address the problem of 3D object recognition from a single 2D image using models database. We propose a method based on geometric quasi-invariant features of the 2D images. We index the 2D images in a model base using a modified quad-tree technique that enhance the research process. The final vote that matches the 2D object image to the 3D object of the database is solved by a vector approximation file which overcomes the difficulties of high dimensionality by following not the data partitioning approach of conventional index methods, but rather as filter based approach.

Improved shape from shading without initial information

The number of constraints imposed on the surface, the light source, the camera model and in particular the initial information makes shape from shading (SFS) very difficult for real applications. There are a considerable number of approaches which require an initial data about the 3D object such as boundary conditions (BC). However, it is difficult to obtain these information for each point of the object Edge in the image, thus the application of these approaches is limited. This paper shows an improvement of the Global View method proposed by Zhu and Shi [1]. The main improvement is that we make the resolution done automatically without any additional information on the 3D object. The method involves four steps. The first step is to determine the singular curves and the relationship between them. In the second step, we generate the global graph, determine the sub-graphs, and determine the partial and global configuration. The proposed method to determine the convexity and the concavity of the singular curves is applied in the third step. Finally, we apply the Fast-Marching method to reconstruct the 3D object. Our approach is successfully tested on some synthetic and real images. Also, the obtained results are compared and discussed with some previous methods.

Tabu Search to Solve the Shape from Shading Ambiguity

The three-dimensional reconstruction of an object from one gray-scale image is a basic problem in computer vision. This problem is known by Shape From Shading. It is considered as an ill-posed due to the ambiguity of the convexity or the concavity of the surface to be reconstructed. A method of solving the ambiguity based on the singular points of the image was proposed in this paper. It uses tabu search to determine the optimal solution of the solution space. The proposed method determines all 3D objects for an image in a short time. The proposed method has been tested on synthetic and real images, it was compared with another method of resolution.

SIMIR: New mean SIFT color multi-clustering image retrieval

Content based image retrieval systems (CBIR) are used to search images on the basis of their visual content in a huge image database. This approach uses a multi-clustering technique with a multi-searching process. In addition it integrate, Mean SIFT (Scale Invariant Feature Transform) descriptor as local feature and HSV(hue, saturation, value) histogram as global feature. Our proposition aims to a maximum separation of the execution to gain time and keep the performance of each descriptor. Local and global features are combined for more relevant results. Getting several views to relevant results to cover the subjectivity in displaying results. This article, detailed our proposed method with a comparison with the FIRE (Flexible Image Retrieval) Engine and LIRE (Lucene Image Retrieval). The results demonstrate the feasibility and relevance of our proposition.