结合HSV空间的改进二维Otsu免疫组化图像分割

IMMUNOHISTOCHEMICAL IMAGE SEGMENTATION USING IMPROVED TWO-DIMENSIONAL OTSU AND COMBINING HSV SPACE

肝脏免疫组化图像中阳性区域的定量分析对肝癌的早期诊断有非常重要的意义。针对真彩色免疫组化图像特征,结合HSV空间对二维Otsu算法进行改进。首先针对二维Otsu算法每次计算类间测度矩阵的迹需要遍历整幅图像导致运算量大耗时多的不足,提出一种快速递推算法,利用快速Otsu算法对图像进行预分割;然后针对分割结果中目标区域包含的少量阴性区域,结合图像的HSV空间特征进行优化。将预分割结果与H分量作交集运算,将交集运算结果与预分割结果作差集运算,得到初分割结果;将初分割结果与H分量和S分量的交集运算结果做并集运算,得到最终分割结果。通过与Otsu的对比实验表明,改进算法更好地实现了阳性区域的目标提取,提高了分割的精度。

The quantitative analysis of the positive area of liver immunohistochemical image is significant to the early diagnosis of liver cancer. Aiming at the features of true colour immunohistochemical image,we improved the two-dimensional Otsu algorithm in combination with HSV space. First,in view of the disadvantage that the two-dimensional Otsu has to traverse entire image when ever to calculate the trace of inter-class measure matrix and in turn leads to heavy computation and large time consumption,we proposed a fast recursion algorithm,which uses the fast Otsu to do the pre-segmentation on the image. Then,aiming at the small amount of negative area contained in target area of the segmented result,we optimised it combining the HSV space feature of the image. We carried out the intersection operation on the presegmentation result and the H component,and the subtraction operation on the intersection operation result and the pre-segmentation result to get initial segmentation result,and then carried out the union operation on the initial segmentation result and the result of H and S components＇ intersection operation,thus obtained the final segmentation result. By the contrast experiment with Otsu,the improved algorithm better realised the target extraction in positive area and improved the segmentation accuracy.