This paper provides a comprehensive survey of technical achievements in image retrieval, especially content-based image retrieval, which has been active and prosperous in recent years. The survey includes 100+ papers covering image feature representation and extraction, multidimensional indexing, and system design, the three fundamental bases of content-based image retrieval. Open research issues and future promising directions are identified based on current technology and real-world applications. Image retrieval has been a very active research area since the 1970s, with contributions from database management and computer vision. Text-based image retrieval, which uses text annotations and database systems, has faced challenges such as labor-intensive annotation and subjective human perception. Content-based image retrieval, which indexes images by their visual content like color, texture, and shape, was proposed to overcome these challenges. This approach has led to many techniques and systems, primarily developed by the computer vision community. The paper discusses various visual features used in image retrieval, including color, texture, shape, and color layout. Color features include histograms, color moments, and color sets. Texture features involve co-occurrence matrices, Tamura texture representation, and wavelet transforms. Shape features include Fourier descriptors, moment invariants, and algebraic curves. Color layout features use local color information and spatial relations. Segmentation techniques are also important for extracting shape and layout features. Efficient multidimensional indexing techniques are needed for large-scale image collections. Dimension reduction methods like KLT and clustering are used to reduce feature dimensions. Multidimensional indexing techniques such as R-trees and k-d trees are explored for efficient retrieval. However, non-Euclidean similarity measures require special handling. Image retrieval systems like QBIC, Virage, RetrievalWare, Photobook, VisualSEEk, WebSEEk, Netra, and MARS are discussed, highlighting their features and performance. These systems support various search options, including search by example, text, and sketch. Future research directions include human-in-the-loop systems, integrating high-level concepts with low-level features, web-oriented image retrieval, high-dimensional indexing, and performance evaluation criteria. Human perception of image content is crucial for developing effective retrieval systems. Establishing standardized testbeds and improving evaluation metrics are also important for advancing image retrieval technology.This paper provides a comprehensive survey of technical achievements in image retrieval, especially content-based image retrieval, which has been active and prosperous in recent years. The survey includes 100+ papers covering image feature representation and extraction, multidimensional indexing, and system design, the three fundamental bases of content-based image retrieval. Open research issues and future promising directions are identified based on current technology and real-world applications. Image retrieval has been a very active research area since the 1970s, with contributions from database management and computer vision. Text-based image retrieval, which uses text annotations and database systems, has faced challenges such as labor-intensive annotation and subjective human perception. Content-based image retrieval, which indexes images by their visual content like color, texture, and shape, was proposed to overcome these challenges. This approach has led to many techniques and systems, primarily developed by the computer vision community. The paper discusses various visual features used in image retrieval, including color, texture, shape, and color layout. Color features include histograms, color moments, and color sets. Texture features involve co-occurrence matrices, Tamura texture representation, and wavelet transforms. Shape features include Fourier descriptors, moment invariants, and algebraic curves. Color layout features use local color information and spatial relations. Segmentation techniques are also important for extracting shape and layout features. Efficient multidimensional indexing techniques are needed for large-scale image collections. Dimension reduction methods like KLT and clustering are used to reduce feature dimensions. Multidimensional indexing techniques such as R-trees and k-d trees are explored for efficient retrieval. However, non-Euclidean similarity measures require special handling. Image retrieval systems like QBIC, Virage, RetrievalWare, Photobook, VisualSEEk, WebSEEk, Netra, and MARS are discussed, highlighting their features and performance. These systems support various search options, including search by example, text, and sketch. Future research directions include human-in-the-loop systems, integrating high-level concepts with low-level features, web-oriented image retrieval, high-dimensional indexing, and performance evaluation criteria. Human perception of image content is crucial for developing effective retrieval systems. Establishing standardized testbeds and improving evaluation metrics are also important for advancing image retrieval technology.