METHOD AND APPARATUS FOR IMAGE STITCHING

Abstract

The disclosure discloses a method and an apparatus for image stitching, the method including: performing stitching feature extraction on target images based on deep-learning neural-network model to obtain stitching features of the target images, selecting a reference image and a to-be-registered image from the target images; calculating a similarity measure value between the first stitching feature on a to-be-stitched side of the reference image and a second stitching features of the to-be-registered image; determining the second stitching feature, the similarity measure value between which and the first stitching feature exceeds a preset value, as a target stitching feature, and taking a position that corresponds to the target stitching feature as a stitching line; and reserving, within an overlapping area of splicing the reference image and the to-be-registered image, a portion of an intermediate image in the reference image and the to-be-registered image corresponding to the overlapping area for splicing the reference image and the to-be-registered image, wherein the overlapping area of splicing the reference image and the to-be-registered image is defined by the stitching line.

Inventors

DING, KUN
YANG, SHUAI
LI, WANGPEI
HUANG, SHENG
ZHUANG, YITANG

Assignees

Shanghai Hanshi Information Technology Co., Ltd.

Dates

Publication Date: 20260513
Application Date: 20240619