Localized Radon transform-based detection of linear features in noisy images

This thesis presents a Radon transform-based approach to the detection of linear features in images characterized by high noise levels. The key element of this technique is a modification to the Radon transform where the intensity integration is performed over short line segments rather than across the entire image. This localized Radon transform is tested on several synthetic images with very high noise levels. The results of this testing demonstrate the superior results that are obtained when the linear features of interest are significantly shorter than the image dimensions. A Unear feature is defined according to its characteristics in both the longitudinal and transverse directions. Based on this definition, a linear feature detection algorithm is developed which utilizes the localized Radon transform. In this algorithm, the transform space is subjected to processing which serves to isolate and locate the responses of linear features and suppress the responses of false alarms. This algorithm is tested on both synthetic images and real images, including Synthetic Aperture Radar (SAR) images of ship wakes, to demonstrate its robustness in the presence of noise.