Sébastien Mavromatis

A hybrid method combining pixel-based and object-oriented methods and its application in Hungary using Chinese HJ-1 satellite images

Pixel-based and object-oriented processing of Chinese HJ-1-A satellite imagery (resolution 30 m) acquired on 23 July 2009 were utilized for classification of a study area in Budapest, Hungary. The pixel-based method (maximum likelihood classifier for pixel-level method (MLCPL)) and two object-oriented methods (maximum likelihood classifier for object-level method (MLCOL) and a hybrid method combining image segmentation with the use of a maximum likelihood classifier at the pixel level (MLCPL)) were compared. An extension of the watershed segmentation method was used in this article. After experimenting, we chose an optimum segmentation scale. Classification results showed that the hybrid method outperformed MLCOL, with an overall accuracy of 90.53%, compared with the overall accuracy of 77.53% for MLCOL. Jeffries–Matusita distance analysis revealed that the hybrid method could maintain spectral separability between different classes, which explained the high classification accuracy in mixed-cover types compared with MLCOL. The classification result of the hybrid model is preferred over MLCPL in geographical or landscape ecological research for its accordance with patches in landscape ecology, and for continuity of results. The hybrid of image segmentation and pixel-based classification provides a new way to classify land-cover types, especially mixed land-cover types, using medium-resolution images on a regional, national, or global basis.

Medical image segmentation using texture directional features

Medical image segmentation can often be performed through tissue texture analysis. One of the most recent and interesting ideas to do that is to take into account the distribution of local maximum orders. We have followed up this idea by using directional maximums and we have applied it to tissue differentiation. Two problems are emerging now: one is the identification of a given texture (labeling) and another one is the characterization of the different areas within images (segmentation). In this paper, we present our new approach for texture representation and analysis, and we point out the advances and problems involved in the image segmentation process.

An explorative study on the proximity of buildings to green spaces in urban areas using remotely sensed imagery

Urban areas are major places where intensive interactions between human and the natural system occur. Urban vegetation is a major component of the urban ecosystem, and urban residents benefit substantially from urban green spaces. To measure urban green spaces, remote sensing is an established tool due to its capability of monitoring urban vegetation quickly and continuously. In this study: (1) a Building’s Proximity to Green spaces Index (BPGI) was proposed as a measure of building’s neighbouring green spaces; (2) LiDAR data and multispectral remotely sensed imagery were used to automatically extract information regarding urban buildings and vegetation; (3) BPGI values for all buildings were calculated based on the extracted data and the proximity and adjacency of buildings to green spaces; and (4) two districts were selected in the study area to examine the relationships between the BPGI and different urban environments. Results showed that the BPGI could be used to evaluate the proximity of residents to green spaces at building level, and there was an obvious disparity of BPGI values and distribution of BPGI values between the two districts due to their different urban functions (i.e., downtown area and residential area). Since buildings are the major places for residents to live, work and entertain, this index may provide an objective tool for evaluating the proximity of residents to neighbouring green spaces. However, it was suggested that proving correlations between the proposed index and human health or environmental amenity would be important in future research for the index to be useful in urban planning.

Earth observation using radar data: an overview of applications and challenges

Abstract The first pictures of the earth were taken from a balloon in the mid-19th century and thus started ‘earth observation’. Aerial missions in the 20th century enabled the build-up of outstanding photographic libraries and then with Landsat-1, the first civilian satellite launched in 1972, digital images of the earth became an operational reality. The main roles of earth observation have become scientific, economic and strategic, and the role of synthetic aperture radar (SAR) is significant in this overall framework. Radar image exploitation has matured and several operational programs regularly use SAR data for input and numerous applications are being further developed. The technological development of interferometry and polarimetry has helped further develop these radar based applications. This paper highlights this role through a description of actual applications and projects, and concludes with a discussion of some challenges for which SAR systems may provide significant assistance.

A new approach for registering remote sensing images from various modalities

Image registration is a major issue in the field of Remote Sensing because it provides a support for integrating information from two or more images into a model that represents our knowledge on a given application. It may be used for comparing the content of two segmented images captured by the same sensor at different times; but it also may be used for extracting and assembling information from images captured by various sensors corresponding to different modalities (optical, radar,). The registration of images from different modalities is a very difficult problem because data representations are different (e.g. vectors for multispectral images and scalar values for radar ones) but also, and especially, because an important part of the information is different from an image to another (e.g. hyperspectral signature and radar response). And precisely, any registration process is based, explicitly or not, on matching the common information in the two images. The problem we are interested in is to develop a generic approach that enables the registration of two images from different modalities when their spatial representations are related by a rigid transformation. This situation often occurs, and it requires a very robust and accurate registration process to provide the spatial correspondence. First, we show that this registration problem between images from different modalities can be reduced to a matching problem between binary images. There are many approaches to tackle this problem, and we give an overview of these approaches. But we have to take into account the specificity of the context in which we have to solve this problem: we must select those points of both images that are associated with the same information, and not the other ones, in order to process the pairing that will lead to the registration parameters. The approach we propose is a Hough-like method that induces a separation between relevant and non-relevant pairings, the Hough space being a representation of the rigid transformation parameters. In order to characterize the relevant items in each image, we propose a new primitive that provides a local representation of patterns in binary images. We give a complete description of this approach and results concerning various types of images to register.

Shallow-water Image Enhancement Using Relative Global Histogram Stretching Based on Adaptive Parameter Acquisition

Light absorption and scattering lead to underwater image showing low contrast, fuzzy, and color cast. To solve these problems presented in various shallow-water images, we propose a simple but effective shallow-water image enhancement method - relative global histogram stretching (RGHS) based on adaptive parameter acquisition. The proposed method consists of two parts: contrast correction and color correction. The contrast correction in RGB color space firstly equalizes G and B channels and then re-distributes each R-G-B channel histogram with dynamic parameters that relate to the intensity distribution of original image and wavelength attenuation of different colors under the water. The bilateral filtering is used to eliminate the effect of noise while still preserving valuable details of the shallow-water image and even enhancing local information of the image. The color correction is performed by stretching the ‘L’ component and modifying ‘a’ and ‘b’ components in CIE-Lab color space. Experimental results demonstrate that the proposed method can achieve better perceptual quality, higher image information entropy, and less noise, compared to the state-of-the-art underwater image enhancement methods.

Using mathematical morphology on LiDAR data to extract information from urban vegetation

Accurate delineation of individual tree crowns in human settlement is of vital importance to decision-making in environmental management. Increasing availability of LiDAR data and applications of mathematical morphology imply a paradigm shift in tree crown delineation. This paper introduces a new approach based on “mathematical morphology on grey-level images” that enables such delineation. We consider a LiDAR data set as a grey-level image in which the indexes are the (x,y) locations on the grid and in which each value is the corresponding height of the point acquired by using the LiDAR sensor (i.e. top of the tree at the (x,y) location). We have applied this approach to a large data set in the frame of a partnership with a Hungarian University, and the results we obtain are closely related to what can be seen on a 3D visualization of the LiDAR data set.

Characterization of Similar Areas of Two 2D Point Clouds

We here present a new approach to characterize similar areas of two 2D point clouds, which is a major issue in Pattern Recognition and Image Analysis.

Mesure de similarité entre sous-parties de nuages de points 2D

This communication focuses on the characterisation of a similarity measure between parts of 2D point clouds. This measure is defined thanks to the use of a general knowledge about real point clouds: they share a large amount of one-dimensional structures. These structures can be represented into a unified manner with a new type of primitives; then, we set the link between the existence of common information between parts of point clouds and the geometric relations of their primitives. Thus, we define a similarity measure that is rotationally invariant, and an algorithm to compute it.

A new geometric invariant to match regions within remote sensing images of different modalities

The use of several images of various modalities has been proved to be useful for solving problems arising in many different applications of remote sensing. The main reason is that each image of a given modality conveys its own part of specific information, which can be integrated into a single model in order to improve our knowledge on a given area. With the large amount of available data, any task of integration must be performed automatically. At the very first stage of an automated integration process, a rather direct problem arises : given a region of interest within a first image, the question is to find out its equivalent within a second image acquired over the same scene but with a different modality. This problem is difficult because the decision to match two regions must rely on the common part of information supported by the two images, even if their modalities are quite different. In this paper, we propose a new method to address this problem.