Multi-view Shape-From-Polarization

We developed a method to estimate the 3D shape of black specular object. We estimated the surface normal of the object by observing the polarization state of the light from multiple view, where the object is illuminated from whole direction. The problem that the solution is not unique if we use only the polarization analysis is solved by using the 3D shape of space carving.
[1] Daisuke Miyazaki, Takuya Shigetomi, Masashi Baba, Ryo Furukawa, Shinsaku Hiura, Naoki Asada, "Polarization-based surface normal estimation of black specular objects from multiple viewpoints," Proc. 3D Imaging, Modeling, Processing, Visualization & Transmission (3DIMPVT), pp. 104-111, 2012.10
[2] Daisuke Miyazaki, Takuya Shigetomi, Masashi Baba, Ryo Furukawa, Shinsaku Hiura, Naoki Asada, "Surface normal estimation of black specular objects from multiview polarization images," SPIE Optical Engineering, 2017.04

Metamerism Art 2015

We developed a fundamental technique for metamerism art. Under the mixed light 1, the paint 1 and the paint 2 are aimed to be same. Under the mixed light 2, the paint 1 and the paint 4 are aimed to be different. In addition, the mixed light 1 and the mixed light 2 are aimed to be same. We use additive mixture model for the mixed light. The proposed algorithm automatically finds the proper weight of each LED from the LED database that causes the metameric phenomena.
[1] Daisuke Miyazaki, Mia Nakamura, Masashi Baba, Ryo Furukawa, Shinsaku Hiura, "Optimization of LED illumination for generating metamerism," Journal of Imaging Science and Technology, 2016.11-12. to appear

Metamerism Art 2014

We developed a fundamental technique for metamerism art. Under the light 1, the mixed paint 1 and the mixed paint 2 are aimed to be same, and the mixed paint 3 and the mixed paint 4 are aimed to be same. In addition, the mixed paint 1 and the mixed paint 2 are aimed to be brighter than the mixed paint 3 and the mixed paint 4. Under the light 2, the mixed paint 1 and the mixed paint 4 are aimed to be same, and the mixed paint 2 and the mixed paint 3 are aimed to be same. In addition, the mixed paint 1 and the mixed paint 4 are aimed to brighter than the mixed paint 2 and the mixed paint 3. We use subtractive mixture model for the mixed paint. The proposed algorithm automatically finds the proper weight of each paint from the oil paints' database that causes the metameric phenomena.
[1] Daisuke Miyazaki, Takafumi Saneshige, Masashi Baba, Ryo Furukawa, Masahito Aoyama, Shinsaku Hiura, "Metamerism-based Shading Illusion," In The 14th IAPR Conference on Machine Vision Applications (MVA 2015), pp. 255-258, 2015.05

Metamerism Art 2013

We developed a fundamental technique for metamerism art. Under the light 1, the mixed paint 1 and the mixed paint 3 are aimed to be same, and the mixed paint 2 and the mixed paint 3 are aimed to be different. Under the light 2, the mixed paint 1 and the mixed paint 3 are aimed to be different, and the mixed paint 2 and the mixed paint 3 are aimed to be same. We use subtractive mixture model for the mixed paint. The proposed algorithm automatically finds the proper weight of each paint from the oil paints' database that causes the metameric phenomena.
[1] Daisuke Miyazaki, Kanami Takahashi, Masashi Baba, Hirooki Aoki, Ryo Furukawa, Masahito Aoyama, Shinsaku Hiura, "Mixing paints for generating metamerism art under 2 lights and 3 object colors," Proc. IEEE International Conference on Computer Vision Workshops (CPCV), pp. 874-882, 2013.12

Polarization-based dehazing

We developed a method to remove the haze in images. Haze is removed by analyzing the polarization of the state using the theory in physics. Haze parameters are estimated by comparing the input images and the reference images.
[1] Daisuke Miyazaki, Daisuke Akiyama, Masashi Baba, Ryo Furukawa, Shinsaku Hiura, Naoki Asada, "Polarization-based dehazing using two reference objects," Proc. IEEE International Conference on Computer Vision Workshops (CPCV), pp. 852-859, 2013.12

3D LoG Sharpening

We have developed a sharpening method of 3D shape. We sharpenedthe 3D shape by applying 3D LoG (Laplacian of Gaussian) filter to the 3D shape. At the preprocessing stage, we have an option toapply 3D bilateral filter in order to remove the noise. In case when we have to compare the edge intensity between two sets of 3D shape, we applied 3D edge filter.
[1] Masanari Yokomizo, Daisuke Miyazaki, Masashi Baba, Ryo Furukawa, Masahito Aoyama, Shinsaku Hiura, Naoki Asada, "Optimization of 3D Shape Sharpening Filter Based on Geometric Statistical Values," Proc. IAPR Workshop on Machine Vision Applications (MVA2013), pp. 260-263, 2013.5

Uncalibrated Photometric Stereo with Laser Range Sensor

We have developed a 3D shape measurement method of opaque objects. We removed the shadow and the specular reflection using the principal component analysis, estimated the light source direction using the singular value decomposition, and estimated the surface normal of the object using the photometric stereo. Shadow and specular reflection are removed by iteratively computing the principal component analysis. Light source direction is estimated from the images and the shape obtained by 3D laser range sensor.
[1] Tomoyuki Kamikawa, Daisuke Miyazaki, Masashi Baba, Ryo Furukawa, Masahito Aoyama, Shinsaku Hiura, Naoki Asada, "High Density Shapes Using Photometric Stereo and Laser Range Sensor under Unknown Light-Source Direction," Proc. IAPR Workshop on Machine Vision Applications (MVA2013), pp. 260-263, 2013.5

Metamerism Art 2011-2012

We have innovated a basic framework of metameric artistic illusion. We reproduced the phenomena that two different paintings can be perceived as a same color under a certain light. The proposed algorithm automatically finds the proper weight of each paint from the oil paints' database that causes the metameric phenomena.
[1] Daisuke Miyazaki, Kazuki Nakamura, Masashi Baba, Ryo Furukawa, Masahito Aoyama, Shinsaku Hiura, Naoki Asada, "A First Introduction to Metamerism Art," SIGGRAPH Asia 2012 Posters, pp. 17:1-17:1, 2012.11-12

Virtual Origami Crane

We developed a method to produce a 3D virtual model of origami crane. We detected which orientation the texture will appear on the crane through the observation of the user's process of folding the crane. We represented the virtual origami crane which is almost the same as the real origami crane drawn by the user in a real world by mapping the texture to the 3D shape.
[1] Daisuke Miyazaki, Saori Kagimoto, Masashi Baba, Naoki Asada, "Creating digital model of origami crane through recognition of origami states from image sequence," SIGGRAPH ASIA 2010 Posters, pp. 19, 2010.12

Shadow Removal
- Hierarchical Graph Cut -

We developed a method to remove the shadow area in an image. We removed the shadow of single image using a hierarchical graph cut which is faster than alpha-expansion method. The users can interactively specify the region where they want to remove the shadow.
[1] Daisuke Miyazaki, Yasuyuki Matsushita, Katsushi Ikeuchi, "Interactive removal of shadows from a single image using hierarchical graph cut," IPSJ Transactions on Computer Vision and Applications, vol. 2, pp. 235-252, 2010.

SVD Photometric Stereo

We developed a method to acquire the 3D shape of opaque objects. We estimated the light source directions using SVD (singular value decomposition), and estimated the surface normal using the photometric stereo. We iteratively computed the SVD so that it will be less sensitive to shadows and specular reflections.
[1] Daisuke Miyazaki, Katsushi Ikeuchi, "Photometric stereo under unknown light sources using robust SVD with missing data," Proc. International Conference on Image Processing, 2010.09

Graph Cut Photometric Stereo

We developed a method to acquire the 3D shape of opaque objects. We improved the photometric stereo using the graph cut method so that it will be less sensitive to shadows and specular reflections.
[1] Daisuke Miyazaki, Katsushi Ikeuchi, "Photometric stereo using graph cut and M-estimation for a virtual tumulus in the presence of highlights and shadows," Proc. Workshop on Applications of Computer Vision in Archaeology, 2010.06

Median Photometric Stereo

We developed a method to acquire the 3D shape of opaque objects. We improved the photometric stereo method so that it will be less sensitive to shadows and specular reflections.
[1] Daisuke Miyazaki, Kenji Hara, Katsushi Ikeuchi, "Median Photometric Stereo as Applied to the Segonko Tumulus and Museum Objects," International Journal of Computer Vision, Vol. 86, No. 2-3, pp. 229-242, 2010.01

Wavelet Texture Method

We developed a method to recreate the appearance of objects. We separated the specular reflection component and the diffuse reflection component by polarizers. We represented the reflection of the object by reflection model. We compressed the image data with wavelet transform.
[1] Daisuke Miyazaki, Takushi Shibata, Katsushi Ikeuchi, "Wavelet-Texture Method: Appearance Compression by Polarization, Parametric Reflection Model, and Daubechies Wavelet," International Journal of Computer Vision, Vol. 86, No. 2-3, pp. 171-191, 2010.01

Sky Polarization

We analyzed a polarization of the sky. Sunlight polarizes when it is reflected at aerosols. We measured the polarization state of clear sky and cloudy sky with fish-eye lens and polarizer.
[1] Daisuke Miyazaki, Mahdi Ammar, Rei Kawakami, Katsushi Ikeuchi, "Estimating Sunlight Polarization Using a Fish-eye Lens," IPSJ Transactions on Computer Vision and Applications, Vol. 1, pp. 288-300, 2009

Inverse Polarization Raytracing

We developed a method to acquire the 3D shape of transparent objects. We observed the polarization state of the light reflected/transmitted at the object, and we obtained the surface shape of transparent objects by solving the inverse problem of polarization raytracing.
[1] Daisuke Miyazaki, Katsushi Ikeuchi, "Shape Estimation of Transparent Objects by Using Inverse Polarization Raytracing," IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI), Vol.29, No.11, pp.2018-2030, 2007.11

Rodin Digital Archiving Project


Rodin is the most celebrated sculptor of the 19th century, who is famous with his works such as The Gates of Hell, The Thinker, and The Burghers of Calais. In this project, we obtained a 3D geometrical model of The Burghers of Calais, which is placed in National Museum of Western Art, Ueno, Tokyo. We also verified the historical fact that Rodin created The Burghers of Calais by reusing the same shape in the different part of this work by using the same cast.
[1] Daisuke Miyazaki, Mawo Kamakura, Tomoaki Higo, Yasuhide Okamoto, Rei Kawakami, Takaaki Shiratori, Akifumi Ikari, Shintaro Ono, Yoshihiro Sato, Mina Oya, Masayuki Tanaka, Katsushi Ikeuchi, Masanori Aoyagi, "3D Digital Archive of the Burghers of Calais," Lecture Notes in Computer Science (Proceedings of the 12th International Conference on Virtual Systems and MultiMedia (VSMM2006)), Vol.4270, pp.399-407, Xi'an, China, 2006.10

PLZT Polarization Camera


We developed a device to acquire the polarization state of the light. We used the material called PLZT which is controllable from the computer, and we obtained the Stokes parameters which represent the polarization state of the light.
[1] Daisuke Miyazaki, Noriyuki Takashima, Akira Yoshida, Eiki Harashima, Katsushi Ikeuchi, "Polarization-based Shape Estimation of Transparent Objects by Using Raytracing and PLZT Camera," in Proceedings of SPIE (Polarization Science and Remote Sensing II, Part of SPIE’s International Symposium on Optics and Photonics 2005), Vol.5888, pp.1-14, San Diego, CA USA, 2005.08 ***Invited talk***

Polarization-based Reflection Parameters Estimation

We developed a method to estimate the reflection parameters of the object surface such as object color and surface roughness. We decomposed the specular reflection component and the diffuse reflection component easily by polarization analysis.
[1] Takushi Shibata, Toru Takahashi, Daisuke Miyazaki, Yoichi Sato, Katsushi Ikeuchi, "Creating Photorealistic Virtual Model with Polarization Based Vision System," in Proceedings of SPIE (Polarization Science and Remote Sensing II, Part of SPIE’s International Symposium on Optics and Photonics 2005), Vol.5888, pp.25-35, San Diego, CA USA, 2005.08

Transparent Surface Modeling
- Shape From Two Polarization Images -

We developed a method to acquire the 3D shape of transparent objects. We observed the polarization state of the light reflected on the object surface from two views, and we applied region segmentation and detection of corresponding points, finally, we obtained the surface normal of transparent objects.
[1] Daisuke Miyazaki, Masataka Kagesawa, Katsushi Ikeuchi, "Transparent Surface Modeling from a Pair of Polarization Images," IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI), Vol.26, No.1, pp.73-82, 2004.01

Polarization-based Inverse Rendering

We obtained the 3D shape of the object by analyzing the images observed from single view through a polarizer. Then, we estimated the illumination distribution from specular reflection image, and we also estimated the reflection parameters such as object color and surface roughness.
[1] Daisuke Miyazaki, Robby T. Tan, Kenji Hara, Katsushi Ikeuchi, "Polarization-based Inverse Rendering from a Single View," in Proceedings of International Conference on Computer Vision (ICCV2003), pp.982-987, Nice, France, 2003.10

Transparent Surface Modeling
- Shape From Thermal Radiation Polarization -

We developed a method to acquire the 3D shape of transparent objects. We obtained a candidate of surface normal by observing the polarization state of reflected light in visible light domain. In addition, we observed the polarization state of thermal radiation in infrared light domain, and finally, we obtained the surface normal of transparent objects.
[1] Daisuke Miyazaki, Megumi Saito, Yoichi Sato, Katsushi Ikeuchi, "Determining surface orientations of transparent objects based on polarization degrees in visible and infrared wavelengths," Journal of Optical Society of America A (JOSA A), Vol. 19, No. 4, pp.687-694, 2002.04

Great Buddha Project


We digitally preserved a Great Buddha of Kamakura. We scanned the buddha from many views by 3D laser sensor, and aligned those range data, and merged into a mesh model. We also synthesized CG image of the buddha whose in the era when it was built, who was covered with gold leaf and settled in a daibutsu-den building.
[1] Daisuke Miyazaki, Takeshi Ooishi, Taku Nishikawa, Ryusuke Sagawa, Ko Nishino, Takashi Tomomatsu, Yutaka Takase, Katsushi Ikeuchi, "The Great Buddha Project: Modelling Cultural Heritage through Observation," in Proceedings of the Sixth International Conference on Virtual Systems and MultiMedia (VSMM2000), pp.138-145, Gifu, Japan, 2000.10 ***Best overall paper***

All publication list


Back to homepage