基于改进型组织特异纹理滤波器的乳腺肿瘤超声图像处理

基于乳腺超声图像的多参数纹理分类实验，改进了Gjenna Sfippel等的自适应纹理滤波器，通过引入模糊函数、增加重叠区域和迭代次数的措施，在减少图像噪声的同时，增强肿瘤与周围正常组织的视觉差别。量化比较乳腺超声图像经该滤波算法和几种常用滤波算法处理前后的的统计特征参量和肿瘤边缘检测的精确率，验证了该算法的有效性和优越性。

Breast Tumor Ultrasound Images Processing Based on an Improved Tissue-specific Texture Filter

An improved algorithm of the tissue - specific adaptive texture filter proposed by Gjenna Stippel et al is presented based on experimental results of the multivariate texture classification for breast tumor ultrasound images. With the introduction of the fuzzy function, and the increase of the overlap area and the iteration number, this improved algorithm can reduce the noise in ultrasound images and provide an enhancement of the visual difference between the tumor and the surrounding healthy tissue, This algorithm＇s performance was quantitatively compared with several other filters in the experiment, The comparison includes statistical characteristic parameters of breast ultrasound images and the precision ratio of the semi - automatic extraction of tumor boundaries respectively before and after the filtration. All results demonstrate the validity and the advantage of the algorithm.