*** README

This folder includes image files used for experiments in following paper.

Chika Inoshita, Yasuhiro Mukaigawa, Yasuyuki Matsushita, Yasushi Yagi, ``Surface Normal Deconvolution: Photometric Stereo for Optically Thick Translucent Objects'', The 13th European Conference on Computer Vision (ECCV2014), Sep., 2014.

Although image files in this dataset are free to use, any publications utilizing this data should cite our paper.



*** FILE NAMES

* CapturedImage/??.png
Captured images in 16 bit png format.

* CapturedImage/CalibMask.png
Mask of black sphere area to estimate direction of incident light.
White indicating sphere area, black indicating background.

* ??.png
Aligned cropped images in 16 bit png format.
We use these images as input in experiment.

* ??_calib.png
Cropped images of black sphere.
Some images are painted by black to focus on brightest specular point.

* ??_CalibMask.png
Mask of black sphere area in cropped sphere image.
Whie indicating sphere area, black indicating background.

* PSF_white.png / PSF_pink.png
Point Spread Functions (PSFs) in each image region for deconvolution.
Images in 16 bit png format.

* PSFMask.png
Show the region of available PSFs.
White: PSF_white.png
Black: PSF_pink.png

* incident_directions.txt
Each row consists of x,y,z coordinates of a unit vector indicating direction of incident light.
The x and y axes span the image plane (x points right, y points up).
The z-axis points into the camera along the optical axis.
