A comparison of affine region detectors is presented, evaluating six different methods for detecting regions that are covariant under affine transformations. These detectors include Harris-Affine, Hessian-Affine, MSER, edge-based region detector (EBR), intensity extremal-based region detector (IBR), and salient region detector. The paper compares their performance on a set of test images under varying imaging conditions, including viewpoint changes, scale changes, illumination, blur, and JPEG compression. The goal is to establish a reference test set and performance software to evaluate future detectors in the same framework. The detectors are compared based on two main criteria: repeatability (how well they detect corresponding regions across different transformations) and distinctiveness (how distinguishable the detected regions are). Repeatability is measured by comparing the overlap between detected regions and ground truth regions, while distinctiveness is assessed using the SIFT descriptor. The results show that the detectors vary in their performance, with some excelling in certain conditions and others in others. For example, MSER and edge-based detectors perform well in structured scenes, while Harris-Affine and Hessian-Affine are effective in textured scenes. The paper also discusses the computational complexity and region size differences among the detectors, noting that larger regions generally have better discriminative power but are more prone to occlusion. The study highlights the importance of considering both the type of scene and the imaging conditions when selecting a region detector. The results provide a benchmark for future research in affine covariant region detection.A comparison of affine region detectors is presented, evaluating six different methods for detecting regions that are covariant under affine transformations. These detectors include Harris-Affine, Hessian-Affine, MSER, edge-based region detector (EBR), intensity extremal-based region detector (IBR), and salient region detector. The paper compares their performance on a set of test images under varying imaging conditions, including viewpoint changes, scale changes, illumination, blur, and JPEG compression. The goal is to establish a reference test set and performance software to evaluate future detectors in the same framework. The detectors are compared based on two main criteria: repeatability (how well they detect corresponding regions across different transformations) and distinctiveness (how distinguishable the detected regions are). Repeatability is measured by comparing the overlap between detected regions and ground truth regions, while distinctiveness is assessed using the SIFT descriptor. The results show that the detectors vary in their performance, with some excelling in certain conditions and others in others. For example, MSER and edge-based detectors perform well in structured scenes, while Harris-Affine and Hessian-Affine are effective in textured scenes. The paper also discusses the computational complexity and region size differences among the detectors, noting that larger regions generally have better discriminative power but are more prone to occlusion. The study highlights the importance of considering both the type of scene and the imaging conditions when selecting a region detector. The results provide a benchmark for future research in affine covariant region detection.