Nowadays, web-scale image search engines (e.g. Google Image Search, Bing Image Search) rely almost purely on surrounding text features. This leads to ambiguous and noisy results. Moreover, most of existing ranking methods for image search often return results according to its relevance with the query, leaving its diversity aside. In order to address these problems, in this paper, we proposed a GDRID (Generating Diverse and Relevant Image results with DivRank) visual search reranking algorithm, which extends the DivRank algorithm to enhance the diversity as well as relevance of the initial search results with visual information. DivRank is based on a reinforced random walk in an information network which can automatically balance the prestige and diversity of the top ranked vertices in a principle way. We evaluate GDRID by using empirical experiments on two popular image data sets, MSRA-MM and Bing reranking data sets. Experimental results outperform existing network-based ranking methods in terms of enhancing diversity in prestige.