A multiscale network for deblurring coal slurry foam images

Deblurring coal slurry foam images is critical for coal slurry flotation condition identification. To improve the quality of images acquired in industrial scenarios of flotation production and to achieve real-time monitoring and fast analysis of foam images, it is necessary to deblur the images to help extract image feature information by machine vision. In this paper, an image deblurring network based on a multiscale architecture is proposed. It is divided into two processes: image degradation and image deblurring. The image degradation process generates a blurred image by adding image blur kernel information to a sharp image, and the image deblurring process puts the generated blurred and sharp images into image pairs for training the sharp image generation network to achieve image deblurring. To enhance foam edge during image deblurring, we design a block structure using a logarithmic difference connection. This structure can help the network focus on the parts of the image that has small greyscale changes during image deblurring and makes the greyscale changes smoother. The experimental results show that the proposed network can achieve better deblurring of foam images, and thus, it lays the foundation for the realization of a fully automatic control and monitoring system for flotation production.