Traffic congestion index (i.e., TCI) is a metric for measuring urban road congestion degree. Since traffic congestion has been one of the major issues in most metropolises, it is a crucial demand to know the TCI of every road segment at every time slot. However, it is a challenging task, because the TCI sensors (e.g., road-side sensors, floating vehicles, etc.) are limited in spatial or temporal dimension, leading to sparse TCI data in the spatial-temporal space. In this paper, we propose a method for estimating the missed TCI data for any road segment at any time slot, based on the sparse TCI data reported by the existing TCI sensors and a variety of open ubiquitous data. First, it extracts various urban features which have a correlation with the TCI data from the open ubiquitous data. Second, it fuses the urban features with the sparse TCI data using a collective matrix factorization algorithm, and collaboratively estimates the missed data. The advantage of our method is that it could adapt to the situation of high TCI data sparsity by incorporating external correlations from open ubiquitous data. We evaluate our method with extensive experiments based on a real-world TCI dataset and four open ubiquitous data sources. The results show the effectiveness of our method.