RGB image definition intelligent enhancement method

By optimizing color camera images using an improved UNet network combined with an attention mechanism, the problem of inconsistent channel focusing caused by chromatic aberration is solved, thus improving image quality and making it suitable for scientific research and industrial inspection.

CN121504748APending Publication Date: 2026-02-10EASY THINKING HANGZHOU TECH CO LTD
0 Cites 0 Cited by

Patent Information

Application Number
CN202511644590.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-11-11
Publication Date
2026-02-10

AI Technical Summary

Technical Problem

Existing color cameras suffer from chromatic aberration due to differences in the focus of light from different channels when collecting light of different wavelengths, which affects image quality and makes them unusable, especially in scientific research and industrial inspection.

Method used

An improved UNet network is introduced, combining positional attention and channel attention mechanisms to optimize single-channel images with poor clarity. The clarity and consistency of each channel of the image are improved by training the dataset.

Benefits of technology

Without increasing hardware costs, it significantly improves the image quality of color cameras and solves the problem of inconsistent multi-channel focusing, making it suitable for scientific research and industrial inspection.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121504748A_ABST
    Figure CN121504748A_ABST
Patent Text Reader

Abstract

The invention discloses an RGB image definition intelligent enhancement method, which comprises the following steps: inputting an RGB image into an improved Unet network, and increasing the number of image channels by an input module in the network to output a feature map F1; a down-sampling module performs down-sampling on the feature map F1 and outputs a feature map F2-FN; inputting the feature maps F1, F2-FN-1 to a position attention sub-module and a channel attention sub-module; the position attention sub-module obtains a feature map Pi; a channel attention sub-module obtains a feature map Ki; the attention fusion module fuses the feature map Pi and the feature map Ki to obtain a fusion map Mi; the up-sampling module outputs a spliced graph; the output module changes the spliced image and outputs an enhanced RGB image; according to the method, the attention mechanism is introduced into the improved UNet network, the single-channel image with poor definition is optimized, the definition of the channel is improved, and the problem of multi-channel focusing inconsistency of the single-frame image of the color camera can be effectively solved.
Need to check novelty before this filing date? Find Prior Art