List of Articles speedup robust features (SURF)

• Open Access Article
  • Abstract Page
  • Full-Text
  
  1 - Matching of Remote Sensing Images Using Improved SURF Detector and Direction-Invariant BRISK Descriptor in the Simulator Environment of Affine Transform Functions
  Fatemeh Khalili Farbod Razzazi Abolfazl Hosseini
  Remote sensing images are often captured by a variety of sensors at different times and with various deviation angles. This makes the matching procedure of image pairs be a challenge. To solve this problem, some algorithms have been proposed to improve this matching. On More

  Remote sensing images are often captured by a variety of sensors at different times and with various deviation angles. This makes the matching procedure of image pairs be a challenge. To solve this problem, some algorithms have been proposed to improve this matching. One of the most popular methods is SURF (Speedup robust features) algorithm, which is somewhat resistant to scale changes, rotation of images, brightness variation, and noise. In addition, the algorithm is suitable for the image deviation angles up to 45 degrees. However, the overlap and proximity of the extracted key points in this algorithm are high and it does not provide a suitable spatial distribution for the key points. This study is looking for a method that is resistant to the changes of affine transformation parameters. We use an IMAS (Image matching by affine simulation) simulator environment, which offers a suitable distribution of key points and can be considered as a solution to more angle differences than SURF. A morphology filter is used to find the boundaries and the edges with more clarity in the images. To reveal the key points, the images centers of mass are employed, which address the main direction of feature points and describe the invariable rotation. In addition, RBRISK (Rotation invariant binary robust invariant scalable key point) descriptor is employed in the algorithm which is temporally stable. The results of the experiments show that the proposed method improved the matching rate in satellite images by about 10% with suitable computational complexity. Manuscript profile