Modal analysis using structural responses identified from high-speed cameras is a challenging task. The problem is that the measured displacements are relatively small (typically deep in the sub-pixel range) and submerged in noise due to the low dynamic range of the camera sensor. A typical approach to determine structural responses from high-speed camera data is the digital image correlation (DIC) method, a general, computationally intensive method for identifying displacements. Without knowing the assumptions of the modal analysis, DIC identifies the displacement in the time domain by minimising the difference between two consecutive regions of interest (ROIs). Optical flow is a method based on the change in intensity in a given pixel due to the change in reflection from a moving surface. The displacement is identified from the change in intensity and the spatial gradient of the intensity of the surface. For small, sub-pixel movements, the relationship between intensity change and displacement is linear, which opens up the possibility of performing the modal analysis directly on the pixel intensity measured by the camera. This research applies the recently introduced Morlet-wave modal method and introduces an experimental modal analysis based on a single pixel with optical flow directly from the pixel intensities and the spatial gradient of the intensity. Furthermore, it is shown that the natural frequencies and damping ratios do not require the spatial gradient. The introduced method was successfully applied to the experimental test case where a pixel-based, full-field modal analysis was performed. The influence of averaging the results from multiple pixels in the modal domain is investigated. Modal identification is compared with the results obtained from the displacements identified with a digital image correlation (DIC) method. The introduced direct pixel-based modal analysis provides a robust and numerically efficient way to a full-field modal analysis.