Image-based pose determination

A method of processing camera-captured imagery, depicting a subject, to estimate a pose between the subject and the camera, the pose being characterized, at least in part, by first and second pose parameters drawn from the group consisting of:

• scale, rotation, X-offset, and Y-offset, the method comprising the acts:
• (a) for each of multiple pixels in the imagery, computing a transformed pixel value based on relationships between a value of the pixel, and values of plural neighboring pixels;
• (b) selecting an ordered sequence of L of said transformed pixel values, corresponding to a constellation of L of said multiple pixels, yielding a query L-tuple of transformed pixel values;
• (c) searching a reference data structure to identify a first reference L-tuple that most closely matches the query L-tuple, the first reference L-tuple having associated therewith:
• (i) the first and second pose parameters.

An image processing method for estimating plural affine parameters with which a tiled image signal, included in artwork printed on a physical object, is depicted in camera imagery captured from the physical object, the method including the acts:

• receiving first and second sets of data, one of said sets comprising reference data, corresponding to the tiled signal in a reference state, the other of said sets comprising query data, corresponding to an excerpt of said camera imagery;
• transforming the first set of data to produce a hundred or more counterpart sets of data, each characterized by a different combination of plural affine parameters including scale and rotation, wherein several of said combinations have the same scale parameter but different rotation parameters, and several others of said combinations have the same rotation parameter, but different scale parameters;
• correlating each of said transformed counterpart sets of first data, with said second set of data, to determine which of said counterpart sets of first data yields the largest correlation value;
• and
• outputting data indicating the combination of affine parameters with which said determined counterpart set of first data was transformed, so that a steganographic digital watermark decoder can have an estimate of the affine parameters needed to decode a watermark payload from the camera-captured imagery.

An image processing method for estimating plural affine parameters with which a tiled image signal, included in artwork printed on a physical object, is depicted in camera imagery captured from the physical object, the method including the acts:

• generating a lattice of a hundred or more data sets, each data set corresponding to the tiled image signal distorted by a particular combination of scale and rotation parameters, the lattice being characterized by a regular sampling of scale parameters within a range of scale parameters, and by a regular sampling of rotation parameters within a range of rotation parameters, wherein each of said data sets corresponds to a respective position in the lattice;
• receiving a set of data corresponding to an excerpt of said camera-captured imagery;
• identifying a data set within said lattice, to which the received set of data most nearly corresponds;
• refining scale and rotation parameters with which said identified data set is characterized, to increase correspondence between the received set of data and the identified set of data;
• and
• outputting data indicating said refined scale and rotation parameters to a steganographic digital watermark decoder, for use in decoding a watermark payload from the camera-captured imagery.