“Image Retrieval using Bag of Features”

Author Name : Mansi M. Dayma

ABSTRACT

Many computer vision problems, such as image classification, video search, robot localization, and texture identification, have employed the Bag of Features (BoF) method. Because of its simplicity, it is quite popular. These methods are based on unordered collections of image descriptors that are then quantized and spatial information is discarded. As a result, BoF-based systems have set new performance standards on popular image classification benchmarks and have achieved scalability breakthroughs in image retrieval. This study examines similar studies that address how to improve and/or implement BoF. Recent methods that reduce quantization errors, enhance feature recognition, and speed up picture retrieval are highlighted. Meanwhile, unresolved issues and fundamental challenges are also raised.The optimal approaches for sampling pictures, defining local image characteristics, and assessing system performance are among those challenges.   This research assessed similar studies that deal with enhancing and/or implementing BoF. The optimal approaches for sampling pictures, defining local image characteristics, and assessing system performance are among those concerns.

Bag of Features, Feature Extraction, Quantization, Clustering, Image Representation, Image Classification, Support Vector Machine are some of the terms used in this paper.