Medical image coloring method based on depth color migration

Abstract

The invention discloses a medical image coloring method based on depth color migration. The method comprises the following steps: randomly initializing an initial image, coloring the medical image byusing a deep neural network in combination with a color migration idea; extracting depth features of the medical image, the reference image and the initial image through a deep neural network; and then constructing content loss, color loss, spatial relationship loss and texture detail loss, calculating the gradient of total loss relative to the pixels of the initial image, and iteratively updatingthe initial image by utilizing the gradient until the initial image is matched with the color of the reference image and the texture structure of the target image at the same time; then reconstructing a color medical image by utilizing the color information and the brightness information of the initial image and the original medical image in the YUV space. According to the invention, the medicalimage with color information can be generated, the texture structure of the original medical image is provided, and the original diagnostic significance is maintained.

Description

technical field [0001] The invention belongs to the technical field of medical image processing, and belongs to colorization of medical images, deep learning and color migration methods. Background technique [0002] In the medical field, with the popularization of medical imaging equipment, many imaging technologies have been brought, such as computed tomography (CT), magnetic resonance imaging (MRI), ultrasonic imaging (UI), etc. Human tissue information provided by medical imaging Used in medical diagnosis and research. Due to imaging technology, most medical images are gray-scale images, and a few are color images; in terms of imaging prices, consumers' consumption of gray imaging is much lower than that of color imaging, so most of the currently used imaging results are is a grayscale image. Color transfer is currently a common method for dealing with natural images, and it has a good effect on natural images. By artificially selecting a suitable color image as a ref...

AI Technical Summary

Problems solved by technology

At present, there are the following problems: 1. Most of the current medical image coloring is based on statistical methods, and there is no complete structural method for using deep color transfer to color medical images; 2. How to ensure the content of the image before and after coloring The structure remains unchanged, and the colored medical image is still diagnostic

Images