Image registration using sliding registration windows

摘要

This disclosure pertains to devices, methods, and computer readable media for performing image registration. A few generalized steps may be used to carry out the image registration techniques described herein: 1) acquiring image data from an image sensor; 2) selecting a pair of overlapping image portions from the acquired image data for registration; 3) determining an area of “maximum energy” in one of the image portions being registered; 4) placing an image registration window over both image portions at the determined location of maximum energy; 5) registering the overlapping image portions using only the image data falling within the image registration windows; and 6) determining, according to one or more metrics, whether the image registration window should be shifted from a current location before registering subsequently acquired image portions.

主权项

1. An image registration method, comprising:
receiving a first image from an image sensor, the first image having a first number of rows, each row having a second number of pixels; receiving a second image from the image sensor, the second image having the first number of rows, each row having the second number of pixels, wherein each row in the second image has a corresponding row in the first image and each pixel in the second image has a corresponding pixel in the first image; identifying a subset of rows in the second image, wherein identifying a subset of rows in the second image comprises identifying a subset of contiguous rows, from among a plurality of subsets of contiguous rows in the second image, having a highest score, wherein each of the plurality of subsets of contiguous rows is determined by a location of an image registration window, wherein the image registration window slides one row at a time down a length of the second image, and wherein a score for each of the plurality of subsets of contiguous rows is based on a first number of pixel values in each row of the subset of contiguous rows; and registering the first image and the second image using only a subset of pixels in each of the identified subset of rows in the second image and the corresponding subset of pixels in a corresponding subset of rows in the first image.