A chroma-based Video Segmentation Ground-truth


These pages describe a Chroma Video Segmentation Ground Truth (cVSG), a corpus of video sequences and segmentation masks created to provide a representative test-set whereby video segmentation algorithms can be quantitatively evaluated and fairly compared. Particularly, ground-truth data have been focused on evaluation of motion-based segmentation masks, as motion seems to be a very common criterion for segmentation within a large number of domains.

Aware of the difficulty of extracting motion segmentation masks manually, chroma based techniques have been used to record foregrounds and backgrounds separately, being later combined to achieve final video sequences and accurate segmentation masks almost automatically. Sequences content has been selected to ensure a final set of representative sequences, for which different complexities have been considered when designing an initial set of scripts. As discussed in [1] overall sequence complexity has been descomposed in a series of local elements, the so called "complexity factors" (thus stressing their influence in the overall complexity). An explanation of the considered complexity factors, the sequence scripts and a low quality preview of the final sequences can all be found in the "Content" section.

Foregrounds and backgrounds have been combined trying to obtain a reasonable degree of realism in the final sequence. However, as the focus of this ground-truth was just evaluation, realism has not always been properly achived.

Recall that this ground-truth is available just for research purposes. Research groups or institutions that have downloaded this database include:

 For more details please refer to the "Ground-truth Downloading" section.

[1]  Fabrizio Tiburzi, Marcos Escudero, Jesús Bescós y José M. Martínez, “A Ground-truth for Motion-based Video-object Segmentation”, IEEE International Conference on Image Processing, ICIP'2008, Workshop on Multimedia Information Retrieval: New Trends and Challenges, pp.17-20, SanDiego (USA), Oct. 2008.