一种色彩校正方法、装置、计算机设备及存储介质
By iteratively optimizing the color matrix using weighted gradient descent, the problems of high computational complexity and poor applicability of the color matrix in existing technologies are solved, achieving more efficient image color correction.
CN116489329BActive Publication Date: 2026-07-17AVIC AVIONICS CO LTD
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- AVIC AVIONICS CO LTD
- Filing Date
- 2022-12-26
- Publication Date
- 2026-07-17
AI Technical Summary
Technical Problem
In existing technologies, the color matrix is calculated using the least squares method, which has problems such as poor applicability and high computational complexity. Furthermore, manual fine-tuning is time-consuming and laborious, and it cannot effectively correct image color differences.
Method used
We use a weighted gradient descent method to iteratively optimize the color matrix under standard light sources. Combined with weight settings, we calculate the color matrix under actual ambient light sources and perform color correction pixel by pixel.
Benefits of technology
It improves the accuracy and applicability of color matrix calculation results, reduces the need for manual adjustments, and achieves better color correction effects.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN116489329B_ABST
Abstract
本发明适用于图像处理技术领域，提供了一种色彩校正方法、装置、计算机设备及存储介质，所述方法包括以下步骤：标定标准光源下的色彩矩阵，所述标准光源下的色彩矩阵用于表征标准色卡中每个色块的标准值和真实值之间的关系；利用带权重的梯度下降法对标准光源下的色彩矩阵进行迭代优化，得到优化后的标准光源下的色彩矩阵；计算实际环境光源下的色彩矩阵；将得到的实际环境光源下的色彩矩阵作用到输入图像上逐像素点进行色彩校正。本发明实施例中，采用带权重的梯度下降法计算色彩矩阵，提高了色彩矩阵计算结果的准确性和适用性，而且增加了权重设置，可以保证色彩矩阵对权重值较大的颜色色块有更好的校正效果。
Need to check novelty before this filing date? Find Prior Art