Research On The Application Of Improved Criminisi Algorithm In Underground Coal Wall Image Inpainting

Affected by the working environment,the coal wall images taken underground in the coal rock may contain unexpected objects such as wires,pipelines,and support nets.They may seriously reduce the image quality and interfere with the extraction of mineral information.And the Criminisi algorithm has problems such as structural discontinuity,structural discontinuity,texture error extension and filling error when repairing images with complex texture structure information.So a piecewise adaptive image inpainting algorithm with an enhanced edge structure is proposed.Firstly,the algorithm introduced the data item based on ISEF and the curvature information of the isolux line into the priority function.The part solved the problem of improper restoration order and edge retention.Then,the algorithm used the local variance feature to identify the edge region,and used the information entropy to identify the texture region and the flat region.So that the algorithm adaptively selected the size of the sample block on the basis of considering the type of region.The part solved the problem of "block effect",and improved the restoration quality.Finally,the paper determined the parameters in the adaptive sample block size by the indepth analysis of the fabric image information distribution map with strong edge structure.Compared with the traditional Criminisi method,the peak signal-to-noise ratio is improved by2.005 averagely and the structural similarity is improved by 2.199 averagely.The results show that the piecewise adaptive Criminisi image inpainting algorithm with an enhanced edge structure has the better restoration effect on object removal,and make the repaired coal wall image structure smoother and the texture more consistent,which is more suitable for the removal and restoration task of obstruction in coal wall images.The pretreatment of occlusion removal is carried out by this method can enhance the applicability of underground coal rock composition analysis,destruction type identification,and pore and fracture structure analysis based on digital image processing.There are 42 figures,4 tables and 79 references in this paper.