고정 헤더 영역
상세 컨텐츠
본문
EUROCAST 2007
J. Isern Gonzalez 외
J. Isern Gonzalez 외
주 내용
f 대신 scale factors를 다룸.
방식 비교
1. McLauchlan - fixed params. scale factors and alignment offset angles only. variable state-dimension filter (VSDF)를 사용. iterative인 듯.
2. Agapito - AC from homographies between each image and reference image (the first one), linear와 iterative 모두 비교.
3. Seo - AC from homographies between each pair of views. skew and distortion known and fixed. minimum 4 images are necessary.
방식 비교
1. McLauchlan - fixed params. scale factors and alignment offset angles only. variable state-dimension filter (VSDF)를 사용. iterative인 듯.
2. Agapito - AC from homographies between each image and reference image (the first one), linear와 iterative 모두 비교.
3. Seo - AC from homographies between each pair of views. skew and distortion known and fixed. minimum 4 images are necessary.
결론
optic center (= principal point인 듯) is the most sensible parameter.
McLauchlan is the most accurate, but intractable if the number of points goes up. 그보다 McLauchlan은 fixed parameters.
computational cost for Seo grows exponentially with the sequence length.
the results are comparable with those provided by a well-known manual calibration method (Batista).
there is no correspondence(?) between the parameter error and the global error (reprojection error인 듯). Agapito는 global error는 작지만 param error는 큰 듯.
McLauchlan is the most accurate, but intractable if the number of points goes up. 그보다 McLauchlan은 fixed parameters.
computational cost for Seo grows exponentially with the sequence length.
the results are comparable with those provided by a well-known manual calibration method (Batista).
there is no correspondence(?) between the parameter error and the global error (reprojection error인 듯). Agapito는 global error는 작지만 param error는 큰 듯.
