1. Palette‐based colour normalization for histopathology images
- Author
-
Shengzhe Shi, Yanyan Wang, Kaikai Zhang, Qingqing Wang, Sheng Liu, and Chun Wang
- Subjects
biomedical imaging ,image processing ,medical image processing ,Photography ,TR1-1050 ,Computer software ,QA76.75-76.765 - Abstract
Abstract With the emergence of computer‐aided diagnostic (CAD) systems, the speed of histopathological image analysis and the accuracy of cancer detection have considerably improved. However, the appearance of histopathological slides can vary depending on the consistency of tissue thickness, stain concentrations, and equipment, thus affecting the judgment of CAD systems. This study proposes a palette‐based colour normalization method for histopathology images is proposed to solve the problem of colour differences between histopathology images. The method builds a graphical user interface based on an improved palette generation algorithm and colour transfer definitions, through which the user can complete the corresponding colour modification to achieve colour normalization. The evaluation of the metrics on histopathological images shows that our method is able to achieve higher image quality and better preservation of structural information of the source image compared with four other colour normalization algorithms. The peak signal‐to‐noise ratio values obtained by the proposed method on two publicly available datasets were 24.1914 and 21.3666, and the structural similarity index matrix values were 0.9871 and 0.9760. The proposed method provides new ideas for the development and design of CAD systems.
- Published
- 2024
- Full Text
- View/download PDF