A measuring method for liquid diffusion coefficients(D) was proposed by an asymmetric liquid-core cylindrical lens combined with an feature extracting method of optical images in real time automatically. Based on the Software Development Kit provided by a CMOS camera used, the application codes were programmed to extract and analyze the features of the brightness and width of a diffusion image. On the image features, the focal position corresponding to a specific liquid layer with a certain refractive index was found out, and the curve of focal positions varied with diffusion time was recorded automatically. Finally, the diffusion coefficient was calculated by using Ficks second law on the acquired experimental data. With the proposed method, the diffusion process of ethylene glycol with pure water was researched at the room temperature (25 ℃) and the measured diffusion coefficient (D) is 1.164×10-5 cm2/s, the relative error is only 0.34% that of the literature value. As compared with traditional direction observation method, the obtained results demonstrate that the method avoids the error caused by reading focal position with human eye, implements measurement automation and is characterized by higher measuring speeds and accuracy, shorter computing time and stable measuring results.