A Novel 2D-to-3D Conversion System Using Edge Information IEEE Transactions on Consumer Electronics 2010 Chao-Chung Cheng Chung-Te li Liang-Gee Chen
Introduction Some approaches that can generate 3D content Time-of-flight depth sensor Triangular stereo vision 3D graph rendering
Introduction How does our brain perceive depth? Monocular cues:one of the major categories for depth perception Motion parallax Binocular cues Monocular Cues involve those cues that exist for a single eye. This is one of the major categories for depth perception, as there are several different monocular cues that help in depth perception 2.A nother important monocular cue for depth is provided by motion. As the world moves, whether that movement is caused by the perceiver or by his surrounding environment, objects in the environment move in distinct, predictable patterns. This concept is known as the motion parallax. 3.Cues that involve both eyes are called binocular cues, and form the other major category for depth cues. These cues exist because of the differential location of the two eyes. The above link explains this phenomenon in more detail.
Monocular cues Interposition (overlapping) Relative Height Familiar Size Texture Gradient Shadow Linear Perspective
Block-Based Region Grouping Depth from Prior Hypothesis Proposed System Block-Based Region Grouping Depth from Prior Hypothesis 3D Image Visualization using Bilateral Filtering and Depth Image-Based Rendering
Proposed 2D-to-3D Conversion System
Block-Based Region Grouping Measure the similarity of neighboring blocks The blocks are segmented into multiple groups by MST
Depth from Prior Hypothesis Use a line detection algorithm[9] to detect the linear perspective of the scene C.-C. Cheng, C.-T. Li, P.-S. Huang, T.-K. Lin, Y.-M. Tsai, and L.-G. Chen, “A block-based 2D-to-3D conversion system with bilateral filter,” in Proc. IEEE Int. Conf. Consumer Electronics, 2009
Depth from Prior Hypothesis Find the corresponding depth map gradients Compute the gravity center of the block group as the depth
3D Image Visualization using Bilateral Filtering and Depth Image-Based Rendering Remove the blocky artifacts by cross bilateral filter Then the depth map is used to generate 3D image by DIBR[3] W.-Y. Chen and Y.-L. Chang and S.-F. Lin and L.-F. Ding and L.-G. Chen, “Efficient depth image based rendering with edge dependent depth filter and interpolation,” in Proc. ICME, pp. 1314-1317, 2005
Analysis of Computational Complexity Analysis of Visual Quality Experiment Result Analysis of Computational Complexity Analysis of Visual Quality
Analysis of Computational Complexity The computational complexity is Larger block size implies shorter computational time but lower depth map quality
Analysis of Visual Quality Better results than the other single view based algorithm especially in vertically standing objects as in 8(b) the images violating the initial hypothesis still generate an acceptable quality as in 8(h) Has little side effect
Analysis of Visual Quality
Analysis of Visual Quality Comparing the depth quality and visual comfort over 4 video data types Videos that captured by a stereoscopic camera Proposed algorithm Previous work of [9] Commercial software of DDD’s TriDef
Analysis of Visual Quality
Conclusion The proposed algorithm uses edge information to group the image into coherent regions. A simple depth hypothesis is determined by the linear perspective of the scene. The algorithm is quality-scalable depending on the block size.