US 12,136,192 B2
Image processing apparatus
Ryo Shinohara, Chiyoda-ku (JP); and Naoto Matoba, Chiyoda-ku (JP)
Assigned to NTT DOCOMO, INC., Chiyoda-ku (JP)
Appl. No. 17/600,261
Filed by NTT DOCOMO, INC., Chiyoda-ku (JP)
PCT Filed May 15, 2020, PCT No. PCT/JP2020/019459
§ 371(c)(1), (2) Date Sep. 30, 2021,
PCT Pub. No. WO2020/230891, PCT Pub. Date Nov. 19, 2020.
Claims priority of application No. 2019-091897 (JP), filed on May 15, 2019.
Prior Publication US 2022/0180491 A1, Jun. 9, 2022
Int. Cl. G06T 5/50 (2006.01); G06N 3/04 (2023.01); G06N 3/045 (2023.01); G06N 3/08 (2023.01); G06N 3/082 (2023.01); G06T 1/20 (2006.01); G06V 10/10 (2022.01)
CPC G06T 5/50 (2013.01) [G06N 3/045 (2023.01); G06N 3/082 (2013.01); G06T 1/20 (2013.01); G06V 10/16 (2022.01)] 7 Claims
OG exemplary drawing
 
1. An image processing apparatus comprising:
processing circuitry configured to
correct input images based on first correction data to generate corrected images, and generate a synthesized image by stitching corrected images together;
determine, as a determination result, whether the corrected images are appropriately stitched together in the synthesized image by using a first trained model having learned whether images are appropriately stitched together in an image in which the images are stitched together;
generate second correction data by supplying the input images to a second trained model having learned relationships between correction data used to correct source images to generate corrected images that are appropriately stitched together and the source images; and
output the synthesized image when the determination result is affirmative and output an updated synthesized image generated by causing the processing circuitry to update the synthesized image based on the second correction data when the determination result is negative; generate pieces of third correction data based on the second correction data;
acquire assessment values corresponding to the pieces of third correction data on a one-to-one basis by supplying each of the pieces of third correction data along with the input images to a third trained model having learned relationships between a set, which includes source images and correction data, and an assessment value indicating appropriateness of stitching of corrected images generated by correcting the source images included in the set using the correction data included in the set; and
select one of the pieces of third correction data corresponding to a maximum assessment value among the assessment values, wherein the processing circuitry is configured to output the updated synthesized image obtained by causing the processing circuitry to generate the updated synthesized image based on the one of the pieces of third correction data when the determination result is negative.