Image reconstruction in which unknown patch is replaced by selected patch
원문보기
IPC분류정보
국가/구분
United States(US) Patent
등록
국제특허분류(IPC7판)
H04N-019/59
H04N-005/21
H04N-007/01
G06T-003/40
H04N-019/14
H04N-019/172
H04N-019/54
출원번호
US-0506009
(2014-09-03)
등록번호
US-10225575
(2019-03-05)
국제출원번호
PCT/JP2014/004511
(2014-09-03)
국제공개번호
WO2016/035107
(2016-03-10)
발명자
/ 주소
Kapik, Lee
Sato, Atsushi
Shibata, Takashi
출원인 / 주소
NEC CORPORATION
인용정보
피인용 횟수 :
0인용 특허 :
12
초록▼
An image processing device according to one of the exemplary aspects of the present invention includes: inferring means for selecting, for each of local unknown patches including a target unknown patch, candidate patches from a plurality of input patches based on similarity to the local unknown patc
An image processing device according to one of the exemplary aspects of the present invention includes: inferring means for selecting, for each of local unknown patches including a target unknown patch, candidate patches from a plurality of input patches based on similarity to the local unknown patch, the local unknown patches being images generated from a part of an unknown image, the plurality of input patches being images generated from a plurality of input images, a subject ID (Identifier) being correlated with the input patches that are generated from an input image to which the subject ID is assigned in the plurality of input image; first score calculation means for calculating a score representing nearness of a candidate patch in the candidate patches to a local unknown patch in the local unknown patches; and patch replacement means for calculating a score summation for the subject ID by summing up scores of the candidate patches being correlated with a same subject ID in the candidate patches of the local unknown patches, and selecting, as a selected patch being used for reconstruction of a reconstruction image, a candidate patch that is correlated with the subject ID for which the score summation is highest from the candidate patches selected for the target unknown patch.
대표청구항▼
1. An image processing device comprising: a memory that stores a set of instructions; andat least one processor configured to execute the set of instructions to:select, for each of local unknown patches, candidate patches from a plurality of input patches based on similarity to the local unknown pat
1. An image processing device comprising: a memory that stores a set of instructions; andat least one processor configured to execute the set of instructions to:select, for each of local unknown patches, candidate patches from a plurality of input patches based on similarity to the local unknown patch, the local unknown patches being images generated from a part of an unknown image, the plurality of input patches being images generated from a plurality of input images, a subject ID (Identifier) being correlated with the input patches that are generated from an input image to which the subject ID is assigned in the plurality of input images;calculate a score representing nearness of a candidate patch in the candidate patches to a local unknown patch in the local unknown patches; andcalculate a score summation for the subject ID by summing up scores of the candidate patches being correlated with a same subject ID in the candidate patches of the local unknown patches, and select for a target unknown patch in the local unknown patches, as a selected patch being used for reconstruction of a reconstruction image, a candidate patch that is correlated with the subject ID for which the score summation is highest from the candidate patches selected for the target unknown patch. 2. The image processing device according to claim 1, wherein the at least one processor is configured to:count, for the subject ID assigned to the candidate patches, a number of the candidate patches nearest to any one of the local unknown patches based on the scores, and re-calculate the scores of the candidate patches based on the numbers for the subject ID assigned to the candidate patches. 3. The image processing device according to claim 1, wherein the at least one processor is configured to:generate the input patches from the input images and the local unknown patches from unknown images by partitioning allowing overlap; andcorrelate an object input patch in the input patches with the subject ID assigned to the input image from which the object input patch is generated. 4. The image processing device according to claim 3, wherein the at least one processor is configured to:normalize the input images so that corresponding feature points are at same positions in the input images;reconstruct the reconstruction image by using the selected patch; andpartition the normalized input images into the input patches so as to be allowed to overlap and partition the unknown image into unknown patches so as to be allowed to overlap, the unknown patches being images in which the local unknown patches are included. 5. The image processing device according to claim 3, wherein the at least one processor is configured to: normalize the input images so that corresponding feature points are at same positions in the input images, and degrade the normalized input images;partition the normalized and degraded input images into the input patches so as to be allowed to overlap, generate a patch pair including one of the input patches and a reconstruction patch which is a part of the normalized input image so that the input patch included in each of the patch pairs corresponds to a degraded image of the reconstruction patch included in the patch pair, and partition the unknown image into unknown patches so as to be allowed to overlap, the unknown patches being images in which the local unknown patches are included; andreconstruct the reconstruction image by using the reconstruction patch included in the patch pair including the candidate patch selected as the selected patch for the target unknown patch. 6. The image processing device according to claim 1, wherein the at least one processor is configured to:receive the plurality of input images;receive the unknown image;store the plurality of input patches in image storage;assign the subject ID to each of the plurality of input images according the identity thereof; andcorrelate the subject ID with each of the plurality of input patches according to the identity of the input image from which the each of the plurality of input patches is generated. 7. An image processing method comprising: selecting, for each of local unknown patches, candidate patches from a plurality of input patches based on similarity to the local unknown patch, the local unknown patches being images generated from a part of an unknown image, the plurality of input patches being images generated from a plurality of input images, a subject ID (Identifier) being correlated with the input patches that are generated from an input image to which the subject ID is assigned in the plurality of input images;calculating a score representing nearness of a candidate patch in the candidate patches to a local unknown patch in the local unknown patches; andcalculating a score summation for the subject ID by summing up scores of the candidate patches being correlated with a same subject ID in the candidate patches of the local unknown patches, and selecting for a target unknown patch in the local unknown patches, as a selected patch being used for reconstruction of a reconstruction image, a candidate patch that is correlated with the subject ID for which the score summation is highest from the candidate patches selected for the target unknown patch. 8. A non-transitory computer readable storage medium storing a program causing a computer to operate as: inferring processing of selecting, for each of local unknown patches, candidate patches from a plurality of input patches based on similarity to the local unknown patch, the local unknown patches being images generated from a part of an unknown image, the plurality of input patches being images generated from a plurality of input images, a subject ID (Identifier) being correlated with the input patches that are generated from an input image to which the subject ID is assigned in the plurality of input images;first score calculation processing of calculating a score representing nearness of a candidate patch in the candidate patches to a local unknown patch in the local unknown patches; andpatch replacement processing of calculating a score summation for the subject ID by summing up scores of the candidate patches being correlated with a same subject ID in the candidate patches of the local unknown patches, and selecting for a target unknown patch in the local unknown patches, as a selected patch being used for reconstruction of a reconstruction image, a candidate patch that is correlated with the subject ID for which the score summation is highest from the candidate patches selected for the target unknown patch. 9. The non-transitory computer readable storage medium according to claim 8, storing a program causing a computer to operate as: second score calculation processing of counting, for the subject ID assigned to the candidate patches, a number of the candidate patches nearest to any one of the local unknown patches based on the scores, and re-calculating the scores of the candidate patches based on the numbers for the subject ID assigned to the candidate patches. 10. The non-transitory computer readable storage medium according to claim 8, storing a program causing a computer to operate as: image partitioning processing of generating the input patches from the input images and the local unknown patches from unknown images by partitioning allowing overlap; andimage ID registration processing of correlating an object input patch in the input patches with the subject ID assigned to the input image from which the object input patch is generated.
연구과제 타임라인
LOADING...
LOADING...
LOADING...
LOADING...
LOADING...
이 특허에 인용된 특허 (12)
Shibata, Takashi; Iketani, Akihiko; Senda, Shuji, Device, method, and computer readable medium for restoring an image.
William T. Freeman ; Egon C. Pasztor ; Baback Moghaddam, Method for inferring scenes from test images and training data using probability propagation in a markov network.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.