In this paper, impacts of noises included in subpixel shifted and overlapping images (multiple low resolution images) are evaluated based on superresolution experiments using iterative back-projection method. Two kinds of noises of radiometric density noise and geometric position noise have been examined in the experiment. The experimental results show the following. 1) Gaussian density noise decreases image resolution in proportion to magnitude of the noise. 2) Salt-and-pepper density noise influences significantly even a small amount. 3) Gaussian position noise should be reduced under the pixel size of a desired high resolution image. 4) Salt-and-pepper position noise does not influence resolution so much, when noise magnitude is pixel size level of a desired high resolution image.
The authors have been concentrating on developing convenient 3D measurement methods using consumer grade digital cameras, and it was concluded that consumer grade digital cameras are expected to become a useful photogrammetric device for the various close range application fields. On the other hand, mobile phone cameras which have 10 Mpixel were appeared on the market in June, 2009. In these circumstances, convenient 3D measurement using mobile phone cameras are enormously expected in various photogrammetric fields instead of consumer grade digital cameras. With this motive, accuracy aspects of mobile phone cameras are investigated in this paper with respect to lens distortion, stability, reliability, robustness and practicability. The calibration tests for 17 mobile phone cameras (1.3M×1, 2M×6, 3M×4, 5M×3, 8M×2, 10M×1) are conducted using test target indoor and practicability is evaluated at outdoor. This paper presents that mobile phone cameras have ability to take the place of consumer grade digital cameras, and develop the market in digital photogrammetric fields.
The present paper proposes an algorithm to automatically extract the edges of buildings from terrestrial LiDAR data for the sake of three-dimensional modeling. Because three-dimensional LiDAR data are difficult to process for the extraction, first, two-dimensional data (rxy,z) is obtained from the original threedimensional data. Next, two gravity points were calculated using several points in the neighbor of the target point in order to reduce the effects of the noise contaminated in the data. Finally, the interior angle of the target point with two gravity points was defined as an indicator for the edges. The results show that the algorithm can function robustly to extract the edges. While the proposed algorithm is based on the two-dimensional processing, it can be applied for the three-dimensional coordinate by merging different scan line results. It is concluded that the proposed algorithm is potential to apply for the three-dimensional modeling of buildings.
In this technical report, we propose an automatic method to extract matching point between cross section data of river channel. The proposed method can be adapted for high water channel effectively, but not for low water channel. We inspect the effectiveness of the proposed method through experiments with survey data, and show a problem and coverage of the proposed method. Finally, we comment on the respects in which the proposed method is improved and on future prospects.