Let the target image have M interest points, T F = T F 1, T F 2, …, T F M and T F m = x, y, and using the Equation (18), we compose the matrix that describes the distance and orientation of points in S T to each other.
Assume that the source image (or reference image) has a set of J interest points as S F = S F 1, S F 2, …, S F j, …, S F J and S F j = x, y, and using Equation (18), we compose the matrix that describes the distance and orientation of points in S F to each other.
The two main directions in QSR are topological reasoning about regions and reasoning about orientations of point configurations.
In the decoding phase, because each image sub-block has slightly different halftone textures due to the orientations of point spread function, it requires a training-based classifier to distinguish the orientations.
The time-to-contact and the surface orientations of points of interest in the real-world are computed by directly combining the linear description of the cortical flow.
r jk is the distance between two interest points, S F j and S F k. θ jk corresponds to the orientation of interest point S F j to S F k.
Besides, whether square defect or rectangular defect, the defect modes change with the orientation of the point defect localizing in the cell only for a bigger defect filling fraction.
That is, a large number of halftone images which are generated by different orientations of the elliptical point spread function are used for training in the frequency domain in advance, until the classifier can distinguish the orientation of from point spread function from each sub-block halftone texture.
This is reasonable since in lower dimensional spaces the shape and orientation of the point clouds must be correctly estimated and taken into account to identify the clusters in a proper way (see Fig. 4).
