Studies in Surveying and Mapping Science (SSMS) Volume 1 Issue 2, June 2013 www.as-se.org/ssms 17 Real-Time Image Matching Using the Trajkovic and Hedley Algorithm Mary Jane Samonte, Rizan Joseph Alcantara Jr., Marvin Evangelista, Kevin Patrick Rabang School of Information Technology, Mapua Institute of Technology Makati City, Philippines mjcsamonte@mapua.edu.ph; rjjdalcantara@mymail.mapua.edu.ph; mpevangelista@mymail.mapua.edu.ph; kporabang@mymail.mapua.edu.ph Abstract One way to acquire information is through surveillance videos. Changes in the video feed can be seen more accurately through image matching. There are several approaches to image matching; and one of these is the SIFT algorithm. Aside from SIFT, Trajkovic and Hedley is another, if not the fastest and most efficient, of the corner detection algorithms. By replacing the keypoint extraction of the SIFT algorithm to the Trajkovic and Hedley corner detection algorithm, the SIFT algorithm will become faster. The study is composed of different phases, including Pre-Processing, Feature Extraction and the Matching Phase. This study has two objectives: to introduce a new method when it comes to image matching, especially with the use of the SIFT algorithm, and the use of other approaches as alternatives used in surveillance and security. Keywords Trajkovic and Hedley; SIFT Algorithm; Keypoint Matching; Corner Detection Introduction “Image matching” is the process of matching two images to see whether they remain the same or if changes have occurred. Matching is also a relevant and fundamental aspect for many problems in computer vision such as scene recognition, motion tracking and others. There are several approaches to image matching, one of which is keypoint matching. Most keypoint matching methods use the SIFT algorithm (Scale Invariant Feature Transform). SIFT is an image processing algorithm which can be used to detect distinct features between two images. The SIFT algorithm has been shown to be a slow, but robust algorithm for image matching. The most obvious drawback to SIFT is the time it takes to compare two images. However, with modifications like efficient and quality keypoint detection, SIFT could be an extremely robust resource for object detection and image matching. The original SIFT algorithm uses Difference-of- Gaussians (DoG) to perform a scale-space search that not only detects features but also estimates their scale. Although several faster implementations of Lowe's approach have been proposed, the approach is inherently resource intensive and therefore, not suitable for real-time execution. The Trajkovic and Hedley corner detection algorithm is one, if not considered to be the fastest and efficient corner detection algorithms. This was developed by Miroslav Trajkovic and Mark Hedley in 1998 with intent to obtain comparable and repeatable rates and localization performance as the most popular and commonly used corner detectors, while requiring minimum computation. A comparison between Trajkovic8 with Harris/Plessey and Moravec operator shows that the Trajkovic8 is appropriate for real time application. The accuracy rating for Trajkovic8 was fair with a speed rating of excellent. The rates used from the lowest to the highest are poor, fair, good and excellent, respectively. The Trajkovic and Hedley, as corner detector, was considered as an alternative method utilizing the same intuition used in the SUSAN operator. For a given point in the image, the variations in brightness along all lines passing through the point are considered. At corners, the variation in brightness will be high for all lines. The repeatability rate of this algorithm is not as high as the Plessey operator, but it is one of the fastest available corner detectors. With this characteristic, it cannot be clearly classified as the other corner detector that has specific features to classify them. The research proposed to implement the Trajkovic and Hedley algorithm (Trajkovic8) as the keypoint detector for the SIFT algorithmincreased the overall efficiency