Stereo Matching Algorithm Based on Wavelet Transform and Joint Selection

Aiming at the fact that the accuracy and speed of the existing local stereo matching method can not meet the requirements of practical application at the same time, this paper proposes a stereo matching algorithm by means of wavelet transform and image features. Firstly, the image is wavelet transformed to obtain high-dimensional information of the image, the image features are combined for matching cost calculation. Secondly, in terms of parallax correction, a joint selection mechanism is proposed. When small-area joint selection is performed, the edge detection is used to refine the selection area, and the object edge information is fully limited as a joint selection area , ensuring that the algorithm can have an exact match also in a depth discontinuity area . After the 2014 dataset test provided by the Middlebury test platform, the algorithm has higher matching accuracy and faster speed than some excellent algorithms such as SSWTAD algorithm and Adapt Weight algorithm .