An efficient approach for sub-image separation from large-scale multi-panel images using dynamic programming

Multi-panel images are increasingly used in research and medical domains for describing complicated situations like results' comparison in paper; or case depiction of a patient by combining all his medical images into a consolidated image. However, the content based image retrieval (CBIR) systems face the issue of performance decline in terms of poor retrieval accuracy because the individual sub-images of the multi-panel images cannot be accessed during the searching process. Representing multi-panel images in the form of sub-images is a necessary step for improving the retrieval accuracy of CBIR systems. The state-of-the-art multi-panel image segmentation approaches use recursive approach for sub-image separation, which detects the location of the sub-lines of a line in the multi-panel image appearing in its sub-images repeatedly. This characteristic of the available approaches makes the CBIR incapable to provide the intended results to the end users in real time. In this work, a line detection-based method using dynamic programming is proposed for sub-image separation, which detects the position of every line in the multi-panel image only once, instead of several times as in the case of state-of-art approaches. We evaluated the proposed method on a subset of the imageCLEFmed 2013 dataset, containing 1050 images belonging to different groups. The experimental results depict the effectiveness of the proposed method in term of generating the results quickly without losing the accuracy as compare to the state-of-the-art approaches.