Texture Acquisition

Textures can either be captured as images describing the appearance of real-world surfaces, or artificially generated based on the captured information. In both cases an appropriate image acquisition method should take place. The selected method differs d

  • PDF / 1,650,543 Bytes
  • 40 Pages / 439.37 x 666.142 pts Page_size
  • 19 Downloads / 202 Views

DOWNLOAD

REPORT


Texture Acquisition

Abstract Textures can either be captured as images describing the appearance of real-world surfaces, or artificially generated based on the captured information. In both cases an appropriate image acquisition method should take place. The selected method differs depending on type of material/surface being captured as well as on complexity of its inherited light transport properties, and finally on a type of intended application where the measured data will be exploited. This chapter will guide the reader through visual texture acquisition at various levels of complexity: from static textures, dynamic textures to more demanding view- and illuminationdependent measurements based on Bidirectional Reflectance Distribution Functions (BRDF) and Spatially Varying BRDFs (SVBRDF). A higher visual accuracy can be achieved by means of methods based on capturing general reflectance fields or its more practically obtainable approximation using Bidirectional Texture Functions (BTF). The chapter also mentions approaches to local light scattering measurements represented by Bidirectional Surface Scattering Reflectance Distribution Function (BSSRDF).

3.1 High Dynamic Range Texture Acquisition As textured real-world surfaces very often exhibit a great dynamic range of luminance between the lightest and the darkest areas of the image, this aspect has to be taken into account. Such great differences are likely to occur when highly specular surfaces are measured. As every digital sensor has a limited dynamic range that it can effectively capture, multiple measurements must be taken with different exposures to cover the entire required width of the spectral dynamic range. A set of techniques dealing with acquisition of high dynamic range (HDR) images has gotten a lot of attention in image processing, photography, and computer graphics [19]. HDR image can nowadays be obtained by relatively simple post-processing of images taken by inexpensive consumer digital cameras allowing a bracketing mode. A dynamic range of measured surface texture can be very high, especially from shadows to highlights that might in same cases have an intensity similar to a light source itself. Using HDR techniques it is possible to capture the full dynamic range of the scene, however, most displays still have a rather limited, low dynamic range (LDR). In the worst case, 8-bits per color channel result in only 256 intensity levels M. Haindl, J. Filip, Visual Texture, Advances in Computer Vision and Pattern Recognition, DOI 10.1007/978-1-4471-4902-6_3, © Springer-Verlag London 2013

23

24

3

Texture Acquisition

only. While mapping from HDR input signal to LDR display media (i.e., tone mapping) has been challenged by the skills of painters and photographs, a number of tone-mapping methods or operators has been developed in image processing and computer graphics. Their goal is to reduce the HDR data range to fit into a limited display range and simultaneously convey maximum realism of details visible to the human eye. The tone-mapping operat