Worldwide, breast cancer continues to be the primary cause of death for women, with ultrasound being a widely utilized imaging technique for detecting breast abnormalities. Tumor segmentation in ultrasound images for breast presents significant challenges due to issues such as low contrast, high levels of shadowing, and poorly defined borders. In this study, we propose an enhanced approach for breast ultrasound image segmentation by integrating an image preprocessing pipeline with a deep learning model based on SegNet. The preprocessing phase involved techniques such as Contrast Limited Adaptive Histogram Equalization (CLAHE), Gaussian blur, and Gabor filtering to address issues of acoustic shadowing, speckle noise, and low contrast, thereby improving the quality of the dataset consisting of 647 breast ultrasound images (BUSI dataset). Following preprocessing, the SegNet model was trained and tested. Performance was evaluated using metrics such as Intersection over Union (IoU), Precision, and Accuracy. The model achieved values of 94.65 for IoU, 79.65 for Precision and 95.57 for Accuracy. Additionally, our proposed method was compared against four recent fully automatic segmentation techniques on the same dataset, demonstrating superior performance across all evaluation metrics.