Underwater wireless optical communication (UWOC) has attracted considerable interest owing to its capability of high data rates and low latency. As a crucial component of UWOC, the transmission characteristics of an underwater channel directly impact the system's performance metrics. However, the existing channel models cannot exactly capture the underwater channel states, thus degrading the observability of channel states. This paper proposes a hybrid-field channel model containing both far-field and near-field path components, in which the signal-dependent shot noise (SDSN) is incorporated as well to accurately describe the underwater channel behavior. Then an improved orthogonal matching pursuit (I-OMP) algorithm that estimates the far-field and near-field path components independently with different transform matrices is developed to obtain the underwater channel state. The performance analyses show that I-OMP can improve the estimation accuracy of underwater channels by iteratively minimizing the mean square error (MSE) and utilizing two different transform matrices, demonstrating the advantage of the proposed I-OMP over the existing methods.