Capturing high-quality ocean surface images plays an important role in lots of visual tasks about ocean such as marine oil spill tracking and Ulva (U.) prolifera detection. In the natural scene, sun shines the surface of ocean directly and causes specular reflection, the images of ocean surface captured by camera often have large number of specular highlight component which has great negative impact to visual tasks of ocean. Different from the samples adopted in current highlight removal method, the specular highlight component is distributed in whole ocean surface images. And in many pixels, the specular reflection component is much larger than the pure diffuse reflection and even goes beyond the dynamic range of brightness that image can store. So if just separating highlight component from the original image, many dark spots will remain in the original image and high quality images cannot be obtained. In this paper, we propose a specular reflection removal method for large-scale ocean surface images captured by UVA. The method proposed in this paper includes two parts: highlight detection and image inpainting in highlight regions. The method based on intensity ratio is used to extract the highlight regions in ocean surface images and then image inpainting method based on FMM can inpaint these regions in a relatively short time. Experimental results show that our method can effectively remove the highlight component and get higher quality images of ocean surface.