【摘要】 针对高强度聚焦超声(HIFU)治疗过程中肿瘤的快速识别与自动定位问题，提出了一种基于SEEDS超像素的自动分割方法，能够为医生快速、准确地定位肿瘤提供帮助，大大减少医生的工作量。该方法基于分裂合并的思想，先利用SEEDS将图像分裂成超像素，再提取超像素的纹理和边界信息，使用纹理边界编码压缩算法完成超像素合并，最后在先验信息的限制下实现肿瘤的自动分割。实验结果表明，该方法可以应对图像质量低、背景复杂等不同情况下的肿瘤分割问题，分割结果比较准确。更多还原

【Abstract】 To solve the problem of rapid recognition and automatic localization of tumors in High Intensity Focused Ultrasound(HIFU) therapy, an automatic segmentation method based on Superpixels Extracted via Energy-Driven Sampling(SEEDS) is proposed, which can help doctors locate the tumor quickly and accurately, and reduce the workload greatly. This segmentation method is based on the split-and-merge framework. Firstly, SEEDS algorithm is used to split the image into many superpixels. Then, the texture and boundary information of superpixels are extracted, and the texture boundary coding compression algorithm is used to complete the merging of superpixels. At last, the automatic segmentation of the tumor can be realized under the limitation of priori information. Experiment results show that the algorithm can deal with the problem of tumor segmentation in different situations such as low image quality and complex background, and get more accurate segmentation results.更多还原