Stereoscopic video generation using motion vector based depth with k-means color segmentation

Abstract

Now a day's 3d video applications becoming more and more popular in our daily life. Although more and more 3d applications are made but they are not still enough quality and comfort viewing. There is rising demand for new algorithms to generate stereoscopic video. Stereoscopic video is an illusion effect of the human eye due to the difference in the perception of the left and right eye images which leads to stereo effect. The purpose of this work is to generate low quality of stereoscopic video from the monoscopic video. Input source video to estimate the depth map based on motion vectors (MV), K-means color quantization and segmentation map to get a stereo video which is also known as 3D video. Later fusion with depth and segmentation map is a step to get the final depth map. Filtering is applied to final depth map to avoid the clipping effect in the next step. Finally to synthesize the left-view and the right-view video using the estimated depth map, we compute the parallax value for each pixel in an video from the estimated depth map.