51.3: An Ultra‐Low‐Cost 2‐D/3‐D Video‐Conversion System

Abstract

In this paper, we propose an ultra low cost 2D‐to‐3D conversion system which generates depth maps using the human visual perception characteristics of luminance and color. The proposed method has two major parts, edge feature‐based global scene depth gradient and texture‐based local depth refinement. The near‐to‐far global scene depth is generated by analyzing the edge complexity on each row. Then the local pixel value, Y, Cb, and Cr component of the video content is used to refine the detail depth value by enhance the texture lighting gradient on object surface. Combining the global depth gradient and local depth refinement, the depth map has a comfortable and vivid quality with very low computational complexity. The algorithm can be easily implemented by only line‐based architecture and integrated into the display chip.

Foreground-based depth map generation for 2D-to-3D conversion

“…For the performance evaluation of the proposed approach, we used Cheng's methods [9], [10] as benchmark methods. They are the minimum spanning tree (MST) segmentationbased depth map generation method [10] and the YCbCr color channel-based depth map generation method [9].…”

“…They are the minimum spanning tree (MST) segmentationbased depth map generation method [10] and the YCbCr color channel-based depth map generation method [9]. For converting to 3D images, the same DIBR algorithm [2] was applied to all methods.…”

“…For generating a depth map for an NOV scene image, we used Cheng's method [9], which assigns depth to each horizontal line considering the information on edge numbers. To enhance the visual effect and local depth refinement, the YCbCr color model is used as a depth cue.…”

“…Then, a cross bilateral filtering (CBF) [8] is used to reduce the blocky artifact. For the NOV scene, we employ Cheng's method [9], which is known to provide a good result. This paper is organized as follows.…”

Perceptual multi-cues 2D-to-3D conversion system

“…Consequently, depth information is a key component for 3D reconstruction. Therefore, a number of 2D-to-3D conversion systems [1]- [6], which automatically estimate the depth information from monoscopic videos, have been proposed. Compared to existing methods, we integrate some depth cues, like motion and edge, by mimicking the depth perception of human visual system.…”