Attitude Estimation for Camera Based on Frequency Domain Analysis

In the field of machine vision, a very important task is to use the visual information to estimate the attitude of camera. The traditional approach is to embed a visual mark in the image´s texture for attitude estimation. However, a visual mark would spoil the unity of the image. This paper introduces a novel attitude estimation method using the spatial frequency domain analysis. In this method, the high frequency marker which is hard to be distinguished by the naked eyes is embedded in the image for attitude estimation. According to the spatial frequency change of the marker, the posture change of camera can be inferred. The result shows that this method is reliable and effective in terms of attitude estimation.