Image processing method, image processing device, electronic equipment and readable storage medium

Abstract

The invention discloses an image processing method and device, electronic equipment and a storage medium. The image processing method comprises the following steps: processing a training figure image by using an initial deep learning network to obtain an output figure template; determining an overlapping area of a preset portrait area of the preset portrait template and an output portrait area of the output portrait template; determining the overall similarity between the output portrait area and a preset portrait area according to the overlapping area; determining global loss according to the overall similarity; and adjusting the initial deep learning network according to the global loss to obtain an adjusted deep learning network. According to the image processing method and device, the electronic equipment and the storage medium, the deep learning network is adjusted according to the global loss, so that the output portrait area of the output portrait template obtained after the image is processed by adjusting the deep learning network is more accurate, other post-processing optimization algorithms do not need to be additionally adopted, and then the calculation amount for obtaining the portrait area is smaller.

Description

technical field [0001] The present application relates to the technical field of image processing, and in particular to an image processing method, an image processing device, electronic equipment, and a computer-readable storage medium. Background technique [0002] In related technologies, portrait segmentation algorithms or soft segmentation techniques (alpha matting) are usually used to matt images to obtain portrait regions. However, the portrait regions obtained by these algorithms have problems such as holes and region segmentation errors. In order to solve these problems, we have to Additional post-processing optimization algorithms such as morphological processing or connected domain processing are added, so that the calculation amount of obtaining the portrait area is larger, the acquisition speed is slower, and the power consumption of the electronic device running the algorithm is increased. Contents of the invention [0003] Embodiments of the present applicat...

AI Technical Summary

Problems solved by technology

[0002] In related technologies, portrait segmentation algorithms or soft segmentation techniques (alpha matting) are usually used to matt images to obtain portrait regions. However, the portrait regions obtained by these algorithms have problems such as holes and region segmentation errors. In order to solve these problems, we have to Additional post-processing optimization algorithms such as morphological processing or connected domain processing are added, so that the calculation of obtaining the portrait area is larger, the acquisition speed is slower, and the power consumption of the electronic device running the algorithm is increased.

Images