OBJECTIVE To establish a calibration analysis approach of medical image color under the circumstance of non-direct natural indoor light, and use it. METHODS TRM (Topology Resolve-Map) calibration method was established in accordance with the color reduction algorithm, with self-made color and grey patches being supervision color control patches; it was applied to correct 410 cases of clinical tongue image. RESULTS The before and after calibrated a * ,b * values of each group s tongue color were significantly different (P＜0.05) except for that of whitish group. Compared with pre-calibration, the dispersion and variation coefficient of post-calibration a * , b * values decreased. The overall consensus rate was 75.5 ％ after using Fisher' s classification method and artificial classification method to analyze 49 corrected typical tongue cases, with the Kappa value of consistency test result of 0. 675 (P＜ 0.05). CONCLUSION The characteristics of L * a * b * in varied tongue colors can be better demonstrated after the calibration with TRM. TRM model approach can significantly reduce the color errors of tongue images under natural indoor light, which demonstrates its good color calibration effect.%目的 建立一种室内非直射自然光条件下的医学图像颜色校正分析方法并初步应用.方法 根据色彩还原算法结合自制彩色、灰度色标作为校正的监督色标建立TRM颜色校正方法,并对410幅舌色图像进行了校正.结果 舌色a*、b*值除淡白舌组外,校正前后各组均有显著性差异(P＜0.05),校正后a*、b*值离散度和变异系数较校正前减小.49例典型舌象校正后Fisher's分类方法与人工分类比较总体一致率75.5%,一致性检验结果Kappa值为0.675 (P＜0.05).结论 不同舌象颜色经拓扑剖分-影射还原法校正后能较好的体现其L*a*b*颜色特征.拓扑剖分-影射还原模型校正方法能明显降低自然光条件下舌色图像的色差,具有良好的颜色校正效果.
展开▼
