Automatic Remote-sensing Image Registration Using SURF

Abstract

Abstract-Image registration is a key, essential element in analysis of Remote sensing images. Registration is critical both for initial processing and for end-user processing of those image products for data fusion, and change detection. This paper focused on the feature-based category of image registration algorithms. Many techniques for the detection and description of images' local characteristics have been proposed to register a set of images without user intervention. However, it is unclear which descriptors are more appropriate. The descriptors should be distinctive and at the same time robust to changes in viewing conditions as well as to errors of the sensor. In our evaluation, we have separated the detector from the descriptor as their performance depends on the interest point detector used. The descriptors are compared according to their recall and runtime efficiency and this deals with several geometric and photometric changes. We also propose an extension to the SURF descriptor and the results show the effectiveness of proposed improvements compared to base SURF method. Furthermore, we observe that the SURF descriptor outperforms the others' descriptors. Finally, based on the test results, we propose an approach to register automatically remotely sensed images.

Robust Automated Image Co-Registration of Optical Multi-Sensor Time Series Data: Database Generation for Multi-Temporal Landslide Detection

“…For accommodating multi-sensor effects during co-registration, feature-based techniques, such as scale-invariant feature transform (SIFT) [19] and speeded-up robust features (SURF) [20], are considered to be more suitable, because these techniques use salient features, such as edges, corners, intersections of linear structures and centroids of distinct geometric objects. These features are expected to be geometrically stable despite the sensor-related variability of the image data [21][22][23][24]. However, in rural mountainous areas, like Southern Kyrgyzstan, such distinct time-invariant features are often scarce and unevenly distributed, which largely increases the likelihood for significant co-registration errors [21,25].…”

Automatic Co-registration of Satellite Time Series via Least Squares Adjustment

“…As this descriptor is quite innovative (2008), it has been scarcely applied to remote sensing images. On the other hand, the technical literature reports some scientific contributions [Teke and Temizel, 2010;Bouchiha and Besbes, 2013] where the method was tested on image pairs and compared to other similar operators (SIFT [Lowe, 2004], PCA-SIFT [Ke and Sukthankar, 2004], GLOH , see for a detailed review of image matching results), obtaining sub-pixel precision and robustness against scale variation, translation, rotation and changes in brightness values. SURF relies on a Hessian matrix-based measure for the detector and the distribution of the first-order Haar wavelet responses for the descriptor [Bay et al, 2008].…”

A New Approach to Satellite Time-series Co-registration for Landslide Monitoring

“…This algorithm implements two main functions: a detector capable of finding interest features which are well characterized with respect to the surrounding background and a descriptor which associates a vector of information to any detected features. Descriptor vectors can be used for matching features between different images without any preliminary information such as seed points or other manual measurements, see Teke and Temizel (2010) and Bouchiha and Besbes (2013). This kind of image matching procedure can be classified as feature-based matching (FBM), see Apollonio et al (2014).…”