Radiological image retrieval technique using multi-resolution texture and shape features

摘要: Medical image analysis plays a very indispensable role in providing the best possible medical support to patient. With rapid advancements modern systems, these digital images are growing exponentially and reside discrete places. These help practitioner understanding problem then suitable treatment. Radiological often found be critical constituent of images. So, health care, manual retrieval visually similar becomes tedious task. To address this issue, we have suggested content-based (CBMIR) system that effectively analyzes image’s primitive visual features. Since radiological gray-scale form, contain rich texture shape features only. novel multi-resolution uses for content analysis. Here, employed modified block difference inverse probability (BDIP) block-level variance local (BVLC) features, respectively. Our proposed scheme variable window size feature extraction strategy maintain co-relation extract more salient Further, used MURA x-ray dataset, which has 40561 captured from 12173 different patients demonstrate scheme’s performance. We also performed compared experiments on Brodatz STex texture, Corel-1K, GHIM-10K natural datasets robustness improvement over other contemporaries.