Image processing method and apparatus and program

摘要

An image processing method for registering first and second images including the same subject is provided. The method includes the repeating the following steps multiple times: a first step including determining feature vectors related to gradients of pixels values at coordinates on the first and second images for each of the coordinates, a second step including calculating for all mutually corresponding coordinates on the first and second images, correlation values each equivalent to a value obtained by an N-th (where N is a natural number) power of an absolute value of an inner product of the feature vectors at the coordinates and determining an evaluation value including an integrated value of the correlation values calculated for each of the coordinates, and a third step including changing the second image in such that the evaluation value becomes larger.

主权项

1. An image processing method for registering first and second images including the same subject, the method comprising repeating the following steps multiple times:
a first step comprising:
determining feature vectors related to gradients of pixels values at coordinates on the first and second images for each of the coordinates;when the size of each of gradient vectors for the pixel values at the coordinates is greater than or equal to a predetermined threshold, determining the gradient vectors each normalized by a size of the gradient vector as the feature vectors at the coordinates; andwhen the size of the gradient vector is less than the predetermined threshold, determining a zero vector as the feature vectors at the coordinates; a second step comprising calculating for all mutually corresponding coordinates on the first and second images, correlation values each equivalent to a value obtained by an N-th (where N is a natural number) power of an absolute value of an inner product of the feature vectors at the coordinates and determining an evaluation value including an integrated value of the correlation values calculated for each of the coordinates; and a third step comprising changing the second image such that the evaluation value becomes larger.