Content Based Image Retrieval - Inspired by Computer Vision & Deep Learning Techniques

Abstract

With the growing popularity of digital services and internet technology billions of people are prone to information sharing and uploading photos. For accurate retrieval of images from huge digital image databases, Content Based Image Retrieval (CBIR) method are emerging as an influential next generation tools, with wide range of applications in fields like criminal investigation, shape recognition, medical diagnosis, remote sensing, digital forensic, radar engineering and robotics. The key challenge of CBIR systems lies in building the “semantic gap” that exists between the differences in the way of perceiving things from basic to complex image features. The intention behind our paper is to provide deeper analysis and impact of continuous advancements in Image Retrieval techniques since its origin. An empirical examination of most popular and upcoming techniques in this survey paved the way in selecting the suitable computer vision and Deep Learning Techniques to improve the performances of retrieval systems. Several Benchmark datasets are considered to validate and study the robustness of the techniques influencing feature extraction and classification.