Method of image restoration of the blank strips of electric imaging logs

The processed data of electric imaging logging can provide important support for the identification of lithology and sedimentary facies of complex reservoirs such as carbonate, glutenite, and volcanic. However, electrical imaging logging cannot achieve full borehole coverage imaging, due to many factors in actual production logging. Corresponding image restoration must be carried out on electric imaging images to obtain full borehole image effects and improve the accuracy of logging data interpretation in the future. This paper through the two image restoration methods under the deep learning model: Edge Connect image restoration method using adversarial edge learning, and Deep Image Prior image restoration method using convolutional neural network (CNN). So that in the actual production logging, verify between the restored electrical imaging logging and the initial electrical imaging logging, and compare the differences between them in the conventional logging curve and resistivity curve. We found that the restore effects of the electrical imaging logging in the actual production logging by the two deep learning methods are better than the initial electrical imaging logging. Among them, the EdgeConnect image restoration method using adversarial edge learning is more suitable for the restoration of complex reservoir stratum images with unique textures and highly heterogeneous than the Deep Image Prior image restoration method using convolutional neural networks.