|
|
Extrakce fototextur pro 3D fotometrickou rekonstrukci
Peroutková, Hedvika ; Buriánek, Jan (advisor) ; Suk, Tomáš (referee)
This thesis deals with the problem of creating textures from photographs, especially for a computer model of buildings. The fact that images are taken from diff erent views and border points of a texture are known is supposed in this thesis. Within the frame of this thesis a system for texture extraction from photographs, searching correspondences in textures and merging textures from di erent views to the fi nal high quality texture suitable for a given model was developed. The problem of registration textures obtained from di fferent images is solved both by feature method based on SIFT descriptors and by method arisen from a deformation model. The implemented methods are also applicable for texturing of huge models in suffi cient quality. We restrict our task to texturing 3D surface models built up from planar faces.
|
|
|
Extrakce fototextur pro 3D fotometrickou rekonstrukci
Peroutková, Hedvika ; Suk, Tomáš (referee) ; Buriánek, Jan (advisor)
This thesis deals with the problem of creating textures from photographs, especially for a computer model of buildings. The fact that images are taken from diff erent views and border points of a texture are known is supposed in this thesis. Within the frame of this thesis a system for texture extraction from photographs, searching correspondences in textures and merging textures from di erent views to the fi nal high quality texture suitable for a given model was developed. The problem of registration textures obtained from di fferent images is solved both by feature method based on SIFT descriptors and by method arisen from a deformation model. The implemented methods are also applicable for texturing of huge models in suffi cient quality. We restrict our task to texturing 3D surface models built up from planar faces.
|
|
|
Graphic filter library implemented in CUDA language
Peroutková, Hedvika ; Šedivá, Zuzana (advisor) ; Sedláček, David (referee)
This thesis deals with the problem of reducing computation time of raster image processing by parallel computing on graphics processing unit. Raster image processing thereby refers to the application of graphic filters, which can be applied in sequence with different settings. This thesis evaluates the suitability of using parallelization on graphic card for raster image adjustments based on multicriterial choice. Filters are implemented for graphics processing unit in CUDA language. Opacity and image merging mode can be set for filters. Various types of filters were implemented like histogram, color, convolution filters and others. Within the frame of this thesis a graphic user interface was developed, which enables filter applying and setting for chosen image, measuring time needed for filter changes and saving final image.
|
guest ::
RSS feed.