Andrea Manno-Kovács

Building Detection From Monocular VHR Images by Integrated Urban Area Knowledge

This letter proposes an approach for building detection from single very high resolution optical satellite images by fusing the knowledge of shadow and urban area information. One of the main contributions of this work is in the integration of urban area information: unlike previous studies, we use such information to substantially revise and improve the initial shadow mask. Additionally, we present an effective way to discriminate dark regions from cast shadows, a task that has continuously been reported to be very difficult. In this letter, we benefit from graph cuts to produce a comprehensive solution for automatic building detection: a flexible multilabel partitioning procedure is proposed, in which the number of optimized classes is automatically selected according to the contents of a scene of interest. The results of the evaluation of 14 demanding test patches confirm the technical merit of the proposed approach, as well as its superiority over three recently developed state-of-the-art methods.

Content based image retrieval using salient orientation histograms

Content-aware image retrieval is a very important topic nowadays, when the amount of digital image data is highly increasing. Existing sketch based image retrieval (SBIR) systems perform at a reduced level on real life images, where background data may distort image descriptors and retrieval results. To avoid this, a preprocessing step is introduced in this paper to distinguish between foreground and background, using integrated saliency detection. To build the descriptor only on the most relevant pixels, orientation feature is extracted at salient Modified Harris for Edges and Corners (MHEC) keypoints using an improved edge map, resulting in a Salient Orientation Histogram (SOH). The proposed SBIR system is also augmented with a segmentation step for object detection. The method is tested on the THUR15000 database, containing random internet images. Image retrieval and object detection both give promising results compared to other state-of-the-art methods.

Wetland mapping by fusion of airborne laser scanning and multi-temporal multispectral satellite imagery

ABSTRACT Wetlands play a major role in Europe’s biodiversity. Despite their importance, wetlands are suffering from constant degradation and loss, therefore, they require constant monitoring. This article presents an automatic method for the mapping and monitoring of wetlands based on the fused processing of laser scans and multispectral satellite imagery, with validations and evaluations performed over an area of Lake Balaton in Hungary. Markov Random Field models have already been shown to successfully integrate various image properties in several remote sensing applications. In this article, we propose the multi-layer fusion Markov Random Field model for classifying wetland areas, built into an automatic classification process that combines multi-temporal multispectral images with a wetland classification reference map derived from airborne laser scanning (ALS) data acquired in an earlier year. Using an ALS-based wetland classification map that relied on a limited amount of ground truthing proved to improve the discrimination of land-cover classes with similar spectral characteristics. Based on the produced classifications, we also present an unsupervised method to track temporal changes of wetland areas by comparing the class labellings of different time layers. During the evaluations, the classification model is validated against manually interpreted independent aerial orthoimages. The results show that the proposed fusion model performs better than solely image-based processing, producing a non-supervised/semi-supervised wetland classification accuracy of 81–93% observed over different years.

Direction selective vector field convolution for contour detection

Active contour methods are widely used for efficient contour detection. This paper proposes a novel contribution for the Harris based Vector Field Convolution (HVFC) method, using the orientation information of feature points in the image by analyzing the gradient information in the small neighborhood. Based on the orientation information, relevant edges are emphasized and an improved edge map is used in the iterative process. The main advantage of the introduced Directional HVFC (DHVFC) method is the ability of exploiting orientation information for increased contour detection accuracy even in case of high curvature boundaries and strong background clutter. The quantitative and qualitative evaluation and comparison with other state-of-the-art methods show that the additional directional information increases the detection performance.

A joint approach of building localization and outline extraction

In this paper we introduce a novel method for building localization and 2D outline extraction in remotely sensed images. A robust Marked Point Process (MPP) model attempts to detect and separate the individual building segments and gives a rough rectangular estimation about the geometry of each entity. The refinement of the detection is achieved by an active contour model, which is initialized by the convex hull of the Harris feature points extracted around the MPP step’s output rectangles. The method is tested in real aerial images provided by the Hungarian Institute of Geodesy, Cartography and Remote Sensing.