Weld quality is crucial for pipeline safety and reliability. The automatic recognition of weld images has developed into a crucial research direction in the welding field, aiming to enhance defect identification and reduce human error. Edge detection plays a pivotal role in weld recognition technology. This paper analyzed the differences between the recognition quality of different weld image edge detection operators, including the Roberts, Sobel, Prewitt, Laplacian–Gauss, and Canny operators. Although small structure elements displayed limited noise reduction ability, they effectively preserved image details, while large structure components exhibited strong denoising capabilities but presented coarse edges. Therefore, a new multi-operator fusion processing method was proposed based on morphological image processing technology. To address the issue of excessive edge features in all directions, including horizontal and vertical, in ray welding image defects, this study combined the Prewitt operator with a square structure element, the Roberts linear structure element, and the Sobel square structure operator, improving the shortcomings of the original operator during edge detection calculation. Multi-operator detection based on morphology was confirmed as beneficial for weld edge identification. The lower edge displayed significantly reduced blackness after weld morphology processing, corrosion, expansion, and edge removal, further confirming that large structural factors exhibited strong denoising ability, while small structural components effectively preserved details. This method substantially improved the accuracy of image edge detection and defect size quantification.