Extraction and Representation of Low-Level Image Features for an Improved CBIR System Using PCA Algorithm

Fast and efficient picture search in huge image databases has gained widespread acceptance in a variety of applications these days. CBIR (content-based image retrieval) is a method of retrieving pictures that is based on automatically determined image attributes. It uses a variety of unique picture feature extraction approaches to find relevant photos. Even if higher level qualities are used to eliminate semantic gaps in the data that may be obtained from visualised information, there is a disparity in how different people understand graphical information, and these semantic variances are difficult to eliminate. The presented ultra-real-time CBIR system is based on low-level characteristics. Lower level qualities such as colour, texture, and shape are extracted using various approaches in this study, and all of the information is recorded in feature vector representation format, which is then combined to build a unique feature vector. Then, using the Euclidean Distance Similarity Metric, these extracted image characteristics are compared to other image attributes. Using accuracy and recall rates, the performance of the proposed approach is evaluated with three current CBIR approaches. When compared to eighteen other ML algorithms, the proposed methodology has reported a greater precision-recall rate and is more efficient.

View article page

Download PDF