Image Compression Model for Power Grid Inspection Guided by Main Frequency Components

How to improve the image compression rate as much as possible while ensuring image accuracy to reduce storage device pressure is an important issue that power grid inspection tasks need to face. Considering that the inspection task focuses on the parts of the image related to power facilities, and the frequency principal components of power facilities are in the lower frequency range, this paper proposes an image compression model based on prediction and attention, guided by the frequency principal components of power facilities. The experimental results show that the compression rate of the proposed model is significantly better than the standard JPEG image compression method.