The integration of optical and radar sensors in a common area is an important task in data fusion as well as other applications. The objective of this study is to perform the feature-based registration of ALOS PALSAR and AVNIR-2 images. The proposed method iteratively minimizes the discrepancy between the feature edges in PALSAR and AVNIR-2 images. A coarse-to-fine strategy is also adopted by applying an image pyramid. The accuracy of feature-based matching has increased from low to high resolution. The test images are synchronous PALSAR and AVNIR-2 images in Japan. The experimental results indicate that one pixel accuracy can be achieved using feature-based matching.