Review on Content-based Image Retrieval Models for Efficient Feature Extraction for Data Analysis

Recent advances in multimedia tools and the explosion of online image archives have prompted numerous studies on efficiently retrieving and managing visual data. All kinds of professionals, from journalists to designers to art historians, have the same problem: finding a certain photograph in a big collection. Image retrieval issues prompted academics to develop new methods for representing and indexing visual information. Content-Based Image Retrieval (CBIR) is a new approach to image retrieval relies on qualities intrinsic to the images themselves. A search is "content-based" looks at the image's actual content. Color, texture, shape, and spatial relationships are image content descriptors. Various real-world computer vision applications use multimedia content analysis and digital images make up the bulk of multimedia data. There has been an exponential increase in the complexity of multimedia items in recent years, particularly photos. Millions of images are published to various repositories such as Instagram, Facebook, and other social media. Computer vision researchers have difficulty finding a relevant image in an archive. The majority of search engines use traditional text-based algorithms rely on captions and metadata to find photos. CBIR, image classification, and image analysis have all seen an outbreak of retrieval activities in the recent two decades. Feature vectors are numerical values to represent high-level visuals in CBIR and image classification models. The image feature set and human visual comprehension have a large gap. This research addresses the gaps between picture feature extraction and human visual perception. This paper will discuss numerous recent advancements in CBIR and picture representation extensively. Analysis of various image retrieval and picture representation models, from low-level feature extraction to contemporary semantic deep learning approaches, has been analyzed. A comprehensive review is conducted on key CBIR and image representation principles and recommendations for additional study in the area.