Reliability Measure of Focus Depth Cue for 2D-to-3D Conversion

Abstract

Recently, interest in 3D technologies has been grown. With the supply of 3D TV, 3D contents are needed a lot, but the number of 3D contents is much smaller than the 2D contents. To overcome this, 2D-to-3D conversion technologies have been proposed. In the 2D-to-3D conversion methods, several depth maps corresponding to depth cues are merged because each depth cue has own limitation. However, the reliability of depth cues have not been studied enough. In this paper, the reliability measure of focus depth cue is proposed. Focus depth cue, one of the most important depth cue, is based on the assumption: focus blur increases with the distance of imaging system. But this assumption does not always hold. Besides, focus blur has limitations. The measure of focus blur is affected by other factors(e.g. motion blur). Considering these fact, we model the reliability measure of focus depth map. In the proposed measure, we classify images to find the nearest region and the existence of salient object in an image. And the proposed measure implies that how much the nearest region is focused. To verify the reliability measure, subjective evaluation is conducted. Finally, an example of modification of recent 2D-to-3D conversion method is suggested by applying the proposed measure. This improves the results of 2D-to-3D conversion.