Panoramic image segmentation method based on high and low frequency enhancement, computer system and medium

The invention provides a panoramic image segmentation method based on high and low frequency enhancement, a computer system and a medium, belongs to the field of deep learning image processing, and aims to solve the problem that a traditional time domain convolutional neural network neglects foreground and background feature frequencies. Comprising the following steps: reading a data set image, and preprocessing the image to obtain four-layer residual features; performing feature dimension adjustment on the four-layer residual features; implementing high-low frequency feature separation to obtain high-frequency features and low-frequency features; performing high-frequency reinforcement on the high-frequency features to obtain high-frequency reinforcement features; performing low-frequency reinforcement on the low-frequency features to obtain low-frequency reinforcement features; and performing segmentation prediction on the high-frequency reinforcement features and the low-frequency reinforcement features re