The hue of the biopsy images in different staining processing may be different because of the inconsistent dye; however, it may cause the annoyance of the image processing. In this study, the color normalization based on the hue warping is developed to make the colors in different images be consistent, so that the same criteria can be applied to these images for
segmentation. The proposed color normalization technique is better than that performed by neural network, and does not need resample color patterns for each image. The hues of renal tubule, major AOI (Area of Interesting) in this study, on different images are almost the same after applying the proposed color normalization method. The k-mean algorithm,
based on the chromaticity, is used for image segmentation, and the obtained results are better than that segmented by the conventional method. The reason why the chromaticity is used as the segmentation criteria is that the chromaticity is intimately related to the way the human visual system perceives color. 腎切片影像常會因為染色劑調配得不一致而於每次染色時會造成切片色調的不同，這會造成後續影像處理工作的困擾。本研究提出以色調直方圖調整為基礎的顏色正規化技術來自動調整不同的切片影像，使其色調一致，進而能以相同條件進行影像分割。這個方法的效能優於類神經網路的方法，同時也可以免除對每張不同色調的影像進行重新取樣的困擾。本研究以分割腎切片上的腎小管部位為主，經顏色正規化之後的結果確實可以讓原先不同色調的腎切片影像之腎小管部位具有相同色調。利用kmean演算法並以色度模型做為分割的依據，其分割效果會優於使用傳統的RGB顏色模型，原因是色度模型所展現出來的顏色與人類視覺系統較為接近，且由實驗結果可以證明具有較佳的分割結果。
