In image processing area, both the shape (geometric aspect) and texture (the pattern present on the surface of the object) can be used to analyze and identify an object in an image. In this sense, this project aims to study and develop new approaches to image analysis. The main focus is the study of methods capable of identifying shapes and textures through the analysis of complexity. Among the methods currently studied (selected based on the similarity of their algorithms and methodology) are:
Deterministic walk: This method uses partially self-repulsive walks to explore the texture (surface of the image) in order to quantify the color transition patterns and, consequently, to identify the image.
Fractal dimension: Based on Fractal Theory, this technique enables us to quantify shapes and textures in images by means of complexity, which is measured based on the level of space occupation/irregularity, in the case of shapes, or in terms of homogeneity/heterogeneity of the texture.
Complex networks: The Complex Networks Theory provides an extensive set of tools for modeling a problem in the form of a graph. From the modeling of an image as a graph, properties of this graph can be used to describe and identify/classify the original image.
In order to explore images in a new manner, and, therefore, extract valuable information from them, this work presents a novel approach, which transforms an image in a dynamic system in gravitational collapse process. This approach enables images to evolve and to present different states, each of which offering a new source of information to be extracted.
In recent years, there has been increasing use of remote sensing images and UAV (Unmanned Aerial Vehicle) in applications involving mapping and urban studies. This is due to the fact that this type of image is a rich source of information about the terrestrial surface, besides having a wide coverage and a low cost. However, these images are the result of complex interactions among the different elements that constitute the city (road system, blocks, lots, and buildings), which hinders the analysis process. Thus, this project intends to study the state of the art in computational methods of image analysis focusing on areas of architecture, urbanism and precision agriculture.
This project uses 3D shape analyses based on fractal dimensions to quantify both inter and intraspecific variations in irregularly shaped organisms such as corals. Compared to traditional methods, fractal dimensions performed at least as good at the interspecific level and considerably better at the intraspecific level.Software to compute Bouligand-Minkowski fractal dimension method from a 3D model (OBJ)
Shape plays an important role in the representation and analysis of objects in images, being one of the most important visual attributes in image processing and computer vision. However, much of the shape information is redundant and can be simplified without impairing the representation of the original object. Polygonal approximation algorithms are based on the elimination of contour points of the shape considered redundant and on the consequent production of a polygon composed of straight segments, thus obtaining a simplified representation of the essence of the shape. Thus, this project aims to study algorithms inspired by nature (i.e., bio-inspired) for the calculation of the polygonal approximation of forms.
Nowadays, Brazil is the world's largest exporter of green coffee, accounting for 30% of world coffee production. Pests and diseases are a very common problem in coffee farms. Among them, rust is a severe disease affecting coffee plantations in Brazil. Rust is caused by an endophytic fungus that attacks the leaf of adult coffee, especially older plants, and if not properly controlled can cause up to 45% reduction in coffee production. In this context, image processing and machine learning techniques can help in the identification process of different plant diseases by speeding up the process or by detecting infections in early stages.
Traditionally, the evaluation of metal microstructures and their physical properties is a subject of study in Metallography. Through microscopy, we obtain images of the microstructures of the material evaluated, while a human expert performs its analysis. However, texture is an important image descriptor as it is directly related to the physical properties of the surface of the object. Thus, in this paper, we propose to use texture analysis methods to automatically classify metal microstructures.
Fibrous dysplasia is a disorder characterized by the replacement of normal bone by dense connective tissue and immature bone trabeculae, commonly found in adolescents and young adults. In craniofacial bones it has a predilection for the maxilla, which can cause severe deformity and asymmetry. The evaluation of fibrous dysplasia on the radiographs of the craniofacial region can be difficult because of the variable appearances and structures that overlap, so that computed tomography is a relevant resource for its correct diagnosis and treatment. The objective of this study was to characterize fibrous dysplasia through lacunarity analysis, a multiscale method to describe patterns of spatial dispersion.
In nature many animals build structures that can be readily measured at the scale of their gross morphology (e.g. length, volume, weight). Capturing individuality such as can be done with the structures designed and built by human architects or artists, however, is more challenging. Here we tested whether computer-aided image texture classification approaches can be used to describe textural variation in the nests of weaverbirds (Ploceus species) in order to attribute nests to the individual weaverbird that built them.
Medical image analysis is a field of intense research, with many approaches being developed over the years. This project proposes to apply texture analysis methods to a relevant medical problem, which consists of classifying cell samples to discover pre-cancerous or cancerous stages.
Colorectal tissue samples for cancer detection
Pap-smear samples for cancer detection
Currently, studies on leaf anatomy have provided an important source of characters helping taxonomic, systematic, and phylogenetic studies. These studies strongly rely on measurements of characters (such as tissue thickness) and qualitative information (structures description, presence–absence of structures). In this work, we used shape and texture analysis methods to extract quantitative data from the plant leaf in order to classify it into its corresponding species.
Analysis of the surface (texture) of the leaf
Leaf shape analysis using complex networks