Limits...
Lightdrum — Portable Light Stage for Accurate BTF Measurement on Site

View Article: PubMed Central - PubMed

ABSTRACT

We propose a miniaturised light stage for measuring the bidirectional reflectance distribution function (BRDF) and the bidirectional texture function (BTF) of surfaces on site in real world application scenarios. The main principle of our lightweight BTF acquisition gantry is a compact hemispherical skeleton with cameras along the meridian and with light emitting diode (LED) modules shining light onto a sample surface. The proposed device is portable and achieves a high speed of measurement while maintaining high degree of accuracy. While the positions of the LEDs are fixed on the hemisphere, the cameras allow us to cover the range of the zenith angle from 0∘ to 75∘ and by rotating the cameras along the axis of the hemisphere we can cover all possible camera directions. This allows us to take measurements with almost the same quality as existing stationary BTF gantries. Two degrees of freedom can be set arbitrarily for measurements and the other two degrees of freedom are fixed, which provides a tradeoff between accuracy of measurements and practical applicability. Assuming that a measured sample is locally flat and spatially accessible, we can set the correct perpendicular direction against the measured sample by means of an auto-collimator prior to measuring. Further, we have designed and used a marker sticker method to allow for the easy rectification and alignment of acquired images during data processing. We show the results of our approach by images rendered for 36 measured material samples.

No MeSH data available.


Step 6 of the algorithm. (a) Input image; (b) mask over the input image; (c) visualization of black/white classification of the masked input image by red and green marks; (d) the images from the homography computed by gradient descent search.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC5375714&req=5

sensors-17-00423-f020: Step 6 of the algorithm. (a) Input image; (b) mask over the input image; (c) visualization of black/white classification of the masked input image by red and green marks; (d) the images from the homography computed by gradient descent search.

Mentions: The four points found in step 5 are used to make a new homography transform that better aligns all 120 averaged images together. The result of this alignment is shown in Figure 20a and is used as the input in the step 6. By yellow colour we denote the region where we have no data from the camera when we apply the currently best known homography transform. We compute the mask for the part of the image where is no radial chequerboard pattern and also we cover the assumed position of the hole with the measured sample. The image with the mask applied, containing only the marker sticker pattern, is shown in Figure 20b. We compute the threshold from the histogram and use a subset of pixels to determine if they are black or white. The subset of 10% of all pixels evaluated is taken at precomputed random positions. Using only a subset of all pixels accelerates this algorithmic step. This is depicted in Figure 20c. The denoted white and black pixels are then fitted against the reference pattern in Figure 14a using a gradient descent search [32] applied to the homography matrix elements (8 degrees of freedom) to find the best match in the image. The result of the improved matching is shown in Figure 20d.


Lightdrum — Portable Light Stage for Accurate BTF Measurement on Site
Step 6 of the algorithm. (a) Input image; (b) mask over the input image; (c) visualization of black/white classification of the masked input image by red and green marks; (d) the images from the homography computed by gradient descent search.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC5375714&req=5

sensors-17-00423-f020: Step 6 of the algorithm. (a) Input image; (b) mask over the input image; (c) visualization of black/white classification of the masked input image by red and green marks; (d) the images from the homography computed by gradient descent search.
Mentions: The four points found in step 5 are used to make a new homography transform that better aligns all 120 averaged images together. The result of this alignment is shown in Figure 20a and is used as the input in the step 6. By yellow colour we denote the region where we have no data from the camera when we apply the currently best known homography transform. We compute the mask for the part of the image where is no radial chequerboard pattern and also we cover the assumed position of the hole with the measured sample. The image with the mask applied, containing only the marker sticker pattern, is shown in Figure 20b. We compute the threshold from the histogram and use a subset of pixels to determine if they are black or white. The subset of 10% of all pixels evaluated is taken at precomputed random positions. Using only a subset of all pixels accelerates this algorithmic step. This is depicted in Figure 20c. The denoted white and black pixels are then fitted against the reference pattern in Figure 14a using a gradient descent search [32] applied to the homography matrix elements (8 degrees of freedom) to find the best match in the image. The result of the improved matching is shown in Figure 20d.

View Article: PubMed Central - PubMed

ABSTRACT

We propose a miniaturised light stage for measuring the bidirectional reflectance distribution function (BRDF) and the bidirectional texture function (BTF) of surfaces on site in real world application scenarios. The main principle of our lightweight BTF acquisition gantry is a compact hemispherical skeleton with cameras along the meridian and with light emitting diode (LED) modules shining light onto a sample surface. The proposed device is portable and achieves a high speed of measurement while maintaining high degree of accuracy. While the positions of the LEDs are fixed on the hemisphere, the cameras allow us to cover the range of the zenith angle from 0∘ to 75∘ and by rotating the cameras along the axis of the hemisphere we can cover all possible camera directions. This allows us to take measurements with almost the same quality as existing stationary BTF gantries. Two degrees of freedom can be set arbitrarily for measurements and the other two degrees of freedom are fixed, which provides a tradeoff between accuracy of measurements and practical applicability. Assuming that a measured sample is locally flat and spatially accessible, we can set the correct perpendicular direction against the measured sample by means of an auto-collimator prior to measuring. Further, we have designed and used a marker sticker method to allow for the easy rectification and alignment of acquired images during data processing. We show the results of our approach by images rendered for 36 measured material samples.

No MeSH data available.