Content-based image retrieval is a technology that is used to identify similar images based on their visual content. Relevant images are found by employing methods that rank images and show the top-ranked images. One important query pertaining to image retrieval methods is regarding as to how to rank the results. This paper proposes a new method based on an unsupervised Hopfield neural network that models human visual memory. In addition, a re-ranking algorithm using post-retrieval analysis is also proposed to refine results by rejecting those top-ranked images that are visually dissimilar. The re-ranking process is based on a combination of spatial information. Results obtained so far indicate that our method is more efficient and promising than other neural network based methods.