Geometric Quality Assessment (GQA) of Earth Observation satellites is crucial for monitoring the system performance and anticipate the accuracy of their products. The Spinning Enhanced Visible and InfraRed Imager (SEVIRI) sensors aboard the geostationary Meteosat satellites of European Organisation for the Exploitation of Meteorological Satellites (EUMETSAT) acquire multispectral images at 5 min (rapid-scan) and 15 min (full-disk scan) intervals in whiskbroom fashion. The existing GQA method used for the SEVIRI images involves matching of landmark points located on the shorelines, which prevents from comprehensive analysis inland. Limitations on detecting further landmarks (keypoints) in the images arise from low spatial resolution (1 and 3 km), large radiometric differences between the spectral bands and low textural content, which is even poorer for night time thermal infrared images. Within the present study initiated by EUMETSAT, a new approach for SEVIRI Level 1.5 imagery was developed for the relative, absolute and interband GQA by using dense image matching with sub-pixel precision. In the study, various image filtering and texture extraction methods were evaluated to obtain dense and reliable keypoints over the full-disk image area. Global image mosaics with superior spatial resolution and accuracy were employed as external reference. 2D shifts between search and reference images were determined the using a sub-pixel image matching technique together with a robust outlier elimination strategy. The results show that the textural information in SEVIRI images can be enhanced greatly by using various filtering methods, such as Laplacian, Sobel, Wallis and Local Binary Patterns. The Laplacian filter was found more suitable for the absolute GQA of infrared band images, which was the most difficult to band to assess. Depending on the cloud cover, thousands of well-distributed keypoints can be detected and employed as independent check points in the absolute GQA process. With the developed GQA Tool, new acquisitions can be assessed rapidly and various accuracy measures on the image geometric quality can also be achieved.