Alignment of digital images and local motion detection for high dynamic range (HDR) imaging

An image processing sub-system, comprising:

• an image capture unit;
• and
• one or more processors coupled to the image capture unit and configured to execute instructions stored in a memory to cause the one or more processors to—
• obtain first and second images of a scene from the image capture unit, wherein the first and second images are temporally offset with respect to one another,
• generate at least a portion of a first image pyramid of down-sampled representations of the first image, wherein the first image forms a base level of the first image pyramid, and wherein each level above the base level comprises a down-sampled version of the first image,
• generate at least a portion of a second image pyramid of down-sampled representations of the second image, wherein the second image forms a base level of the second image pyramid, and wherein each level above the base level comprises a down-sampled version of the second image,
• operate on the first and second image pyramids beginning at the top-most level to determine a combined alignment vector based, at least in part, on a difference bitmap for a current level, wherein the difference bitmap includes an active area having a plurality of tiles, and wherein the instructions to cause the one or more processors to operate comprise instructions to cause the one or more processors to—
• select a pixel shift operation from a plurality of pixel shift operations that, when applied to the difference bitmap yields a smallest shift value,
• determine an alignment vector for the current level based, at least in part, on the smallest shift value,
• prune, when the current level is outside a top-most x levels, one or more tiles from the active area based, at least in part, on a pruning threshold, wherein the active area is determined based at least partially upon the locations of un-pruned tiles from a previous level,
• combine the alignment vectors from each level to produce the combined alignment vector, and
• align the first and second images based, at least in part, on the combined alignment vector.

An image processing method, comprising:

• obtaining first and second images of a scene, wherein the first and second images are temporally offset with respect to one another;
• generating at least a portion of a first image pyramid of down-sampled representations of the first image, wherein the first image forms a base level of the first image pyramid, and wherein each level above the base level comprises a down-sampled version of the first image;
• generating at least a portion of a second image pyramid of down-sampled representations of the second image, wherein the second image forms a base level of the second image pyramid, and wherein each level above the base level comprises a down-sampled version of the second image;
• operating on the first and second image pyramids beginning at the top-most level to determine a combined alignment vector based, at least in part, on a difference bitmap for a current level, wherein the difference bitmap includes an active area having a plurality of tiles, and wherein the act of operating comprises—
• selecting a pixel shift operation from a plurality of pixel shift operations that, when applied to the difference bitmap yields a smallest shift value,
• determining an alignment vector for the current level based, at least in part, on the smallest shift value,
• prune, when the current level is outside a top-most x levels, one or more tiles from the active area based, at least in part, on a pruning threshold, wherein the active area is determined based at least partially upon the locations of un-pruned tiles from a previous level, and
• combining the alignment vectors from each level to produce the combined alignment vector;
• and
• aligning the first and second images based, at least in part, on the combined alignment vector.