In the field of structural health monitoring, structural dynamic properties such as natural frequency, damping ratio, and mode shapes are vital for assessing structural integrity. While accelerometers are widely used, their nature of single-point measurements often provide incomplete dynamic information. Camera-based methods, on the other hand, face challenges in capturing small motions due to camera resolution and the high stiffness of the structure. Phase-based motion evaluation (PME) algorithms aim to address this limitation in detecting subtle motions. However, a PME-based method for extracting structural motion in a random direction (directional motions) and determining the dominant motions under ambient excitation has not been fully explored. This paper introduces an enhanced phase-based evaluation method that employs a 2D Hilbert transform with an angle-steerable design and a novel angular filter, referred to as AS-2DHPME. This approach allows the extraction of motions (i.e., structural vibrations) from various angles. By evaluating the signal-to-noise ratio (SNR) of the acquired signals, the vibration direction is determined, leading to a more accurate extraction of the actual structural vibration modes. In this study, experimental results on a beam structure demonstrates its feasibility and effectiveness in identifying the actual dominant vibrations.