In the underwater environment, conventional hyperspectral imagers for imaging target scenes usually require stable carrying platforms for completing push sweep or complex optical components for beam splitting in long gaze imaging, which limits the system’s efficiency. In this paper, we put forward a novel underwater hyperspectral imaging system inspired by the visual features of typical cephalopods. We designed a visual bionic lens which enlarged the chromatic blur effect to further ensure that the system obtained blur images with high discrimination of different bands. Then, chromatic blur datasets were collected underwater to complete network training for hyperspectral image reconstruction. Based on the trained model, our system only required three frames of chromatic blur images as input to effectively reconstruct spectral images of 30 bands in the working light range from 430 nm to 720 nm. The results showed that the proposed hyperspectral imaging system exhibited good spectral imaging potential. Moreover, compared with the traditional gaze imaging, when obtaining similar hyperspectral images, the data sampling rate in the proposed system was reduced by 90%, and the exposure time of required images was only about 2.1 ms, reduced by 99.98%, which can greatly expand its practical application range. This experimental study illustrates the potential of chromatic blur vision for underwater hyperspectral imaging, which can provide rapid response in the recognition task of some underwater dynamic scenarios.