In terms of the large proportion and multidirectional rotation of small ship targets in remote sensing images, a small ship target detection method based on texture and color enhancement is proposed. Firstly, a generative adversarial network is designed to enhance the texture features of small ship targets and generate high-resolution ship images. Secondly, the deep reinforcement learning algorithm is used to improve the image color, which solves the problem of the low contrast between the ship target and the background color. Thirdly, an adaptive transform feature pyramid network is designed to enhance the global receptive field and effectively deal with the hard extraction of small target features, which is caused by the lack of spatial information in the deep network. Finally, the feature refinement module and circular smooth label are utilized to align feature points and achieve angle regression in a ship target bounding box, which effectively increases the accuracy of detecting ship targets with multidirectional rotation. In addition, related tests are carried out on the public data sets of HRSC2016 and DOTA, and results show that the proposed method achieves an mean average precision of 72.87% and 89.91%, respectively, which is better than the existing mainstream small ship target detection methods.