Multimodal Markov Random Field for Image Reranking Based on Relevance Feedback

Figure 2

Schematic diagram for the proposed multimodal MRF. The MRF takes as input the list of images retrieved by an image retrieval system (IRS), relevance feedback information, and multimodal (textual and visual) features extracted from the images in the list. The structure of the MRF is depicted in Figure 3. The output of the model is a list of reranked images.