Convolutional cost aggregation for robust stereo matching

Jeong, Somi; Kim, Seungryong; Ham, Bumsub; Sohn, Kwanghoon

doi:10.1109/icip.2017.8296737

Cited by 2 publications

(2 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…All of these methods based on hand-designed functions and were unable to capture useful context and semantic information, which leaded to their limited performance. Jeong et al [23] used CNN to learn the convolution kernel for cost aggregation. However, this method need to combine with edge detection task and global energy minimization to achieve a better result.…”

Section: Matching Cost Computationmentioning

confidence: 99%

“…This testifies that parallelly implementing two sizes of receptive fields can make the best use of them and balance the trade-off between the increase of receptive field and the loss of detail. Experimental results of [23] have proved that the conventional hand-crafted aggregation method CBCA [4] outperforms the other traditional aggregation methods such as BF [3], GF [31] and DT [32]. So we compare the performance of our multi-dimension matching cost aggregation network to CBCA in this section.…”

Section: Matching Cost Computation Evaluationmentioning

confidence: 99%

See 1 more Smart Citation

Cascaded Multi-scale and Multi-dimension Convolutional Neural Network for Stereo Matching

Hai

Zhang

et al. 2018

2018 IEEE Visual Communications and Image Processing (VCIP)

View full text Add to dashboard Cite

Convolutional neural networks(CNN) have been shown to perform better than the conventional stereo algorithms for stereo estimation. Numerous efforts focus on the pixel-wise matching cost computation, which is the important building block for many start-of-the-art algorithms. However, those architectures are limited to small and single scale receptive fields and use traditional methods for cost aggregation or even ignore cost aggregation. Differently we take them both into consideration. Firstly, we propose a new multi-scale matching cost computation sub-network, in which two different sizes of receptive fields are implemented parallelly. In this way, the network can make the best use of both variants and balance the trade-off between the increase of receptive field and the loss of detail. Furthermore, we show that our multi-dimension aggregation sub-network which containing 2D convolution and 3D convolution operations can provide rich context and semantic information for estimating an accurate initial disparity. Finally, experiments on challenging stereo benchmark KITTI demonstrate that the proposed method can achieve competitive results even without any additional post-processing.

show abstract

Section: Matching Cost Computationmentioning

confidence: 99%