Open Access

Table 2

Advantages and disadvantages of Stereo image matching algorithms.

Stereo image matching algorithms References Advantages Disadvantages
ADCensus Mei et al. [69] High matching speed, high accuracy The matching fuzziness in duplicate area and similar texture area is easy to cause mismatching
PatchMatch Bleyer et al. [70] Global matching is realized in the inclined plane and sub-pixel matching accuracy is obtained Many operations need to be processed by a single pixel one by one, resulting in slow running speed and need to be carried out in parallel
MatchNet Han et al. [71] A new deep learning network structure with fewer descriptors is proposed, which significantly improves the patch-matching effect Image blocks can only be processed after sampling, and it is impossible to find and match the whole image
Fast Bilateral Solver Barron et al. [72], Barron and Poole [73] Matching speed is very fast The result is easily affected by the reference image
MC-CNN Žbontar and LeCun [74, 76], Ye et al. [77] The deep learning theory is applied to stereo matching for the first time, and the matching accuracy is improved The matching effect is not good in occluded areas, untextured areas and repetitive pattern areas
GA-Net Zhang et al. [78] The matching accuracy of occluded area, untextured area and reflection area is improved Memory usage are too high

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.