Producing Realistic Building Texture Using Video Mosaic

摘要

For realistic building modeling, texture mapping via video sequences is an effective approach. The objective of this study is to semi-automatically produce photo-realistic texture images of building models using video sequences. The textures are generated using mosaic of frames extracted from video sequences and with 40% to 60% of overlaps. The overlapped video frames are merged together based on interest points identified from intersected line features of video images. These interest points are then matched to each other and used to register the images for mosaicing. The generated texture mosaic can then be mapped onto building models.