【染色归一化】DeepPATH中所用方法

最新推荐文章于 2024-07-23 20:40:26 发布

LansinBlog

最新推荐文章于 2024-07-23 20:40:26 发布

阅读量54

点赞数 1

分类专栏：病理图像WSI 文章标签： python 染色归一化

本文链接：https://blog.csdn.net/weixin_45674669/article/details/140607673

版权

病理图像WSI 专栏收录该内容

3 篇文章 0 订阅

订阅专栏

tile = Image.fromarray(self.normalize_tile(tile, Normalize).astype('uint8'),'RGB')

def normalize_tile(self, tile, NormVec):
    Lab = color.rgb2lab(tile)
    TileMean = [0,0,0]
    TileStd = [1,1,1]
    newMean = NormVec[0:3] 
    newStd = NormVec[3:6]
    for i in range(3):
        TileMean[i] = np.mean(Lab[:,:,i])
        TileStd[i] = np.std(Lab[:,:,i])
        # print("mean/std chanel " + str(i) + ": " + str(TileMean[i]) + " / " + str(TileStd[i]))
        tmp = ((Lab[:,:,i] - TileMean[i]) * (newStd[i] / TileStd[i])) + newMean[i]
        if i == 0:
            tmp[tmp<0] = 0 
            tmp[tmp>100] = 100 
            Lab[:,:,i] = tmp
        else:
            tmp[tmp<-128] = 128 
            tmp[tmp>127] = 127 
            Lab[:,:,i] = tmp
    tile = self.Lab_to_RGB(Lab)
    return tile