Otherness feature extraction method for underground image based on Shearlet transform

For the problem that face images collected underground are susceptible to dust interference and most feature extraction methods are sensitive to noise, an otherness feature extraction method for underground image based on Shearlet transform was proposed. First, Shearlet transform was used for multi-directional and multi-scale image decomposition; then each directional sub-graphs with the same scale were encoded and fused, and furthermore fusion was conducted by giving different weights according to Shannon entropy of the subgraphs; finally, Shearlet inverse transform was used to reconstruct otherness image of low-frequency sub-graph and fused high-frequency sub-graph. The experimental results show that the method has a good objective and subjective evaluation.