Image Retrieval Using Low Level and Local Features Contents: A Comprehensive Review

Table 5

Summary of feature fusion-based image retrieval techniques.

Year	Method	Similarity measure	Dataset	Performance measure (%)
2017	A fusion of color moments and seven invariant moments [67]	Euclidean distance	Wang	Precision: 66.2
2017	A fusion of HSV color moments and the Gabor filter-based texture features [68]	Euclidean distance	Wang	Precision: 65.6
2017	A fusion of color histogram features and multilevel Haar wavelet-based texture features [69]	Euclidean distance	Wang	Objective computations not given
2019	Color volume histograms using quantization of HSV color space and edges [70]	L1 distance	Corel-5000	Precision: 60.13
				Recall: 7.21
			Corel-10000	Precision: 48.58
				Recall:5.83
2015	Fusion of SIFT with BoVW [71]	L2 distance	ALOI	Precision: 88
				Recall: 29
			Flickr	Precision: 78
				Recall: 26
2017	A fusion of color histogram features and shape features using a Canny edge detector [72]	Threshold-based	Corel-1K	Precision: 88.2
				Recall:70.02
2018	Chromaticity color moments fused with statistical features of color co-occurrence [73]	Euclidean distance (weighted)	Wang	Precision: 83.83
				Recall: 10.1
2018	SURF descriptors fused with HoG feature descriptors [75]	Euclidean distance	Corel-1000	Precision: 80.61
			Corel-1500	Precision: 76.28; Recall: 15.25
			Corel-5000	Precision: 60.60
				Recall: 12.12
			Caltech-256	Precision: 46.30
				Recall: 09.26
2018	Fusion of the SIFT descriptor with BoVW [76]	Euclidean distance	Corel-1000	Precision: 87.85
				Recall: 17.37
			Corel-1500	Precision: 84.38 Recall: 16.88
2014	A fusion of color features and curvelet features [77]	Euclidean distance	Corel-1000	Precision: 81
2018	SURF- and FREAK-fused feature descriptors using BoVW [74]	Euclidean distance	Corel-1000	Precision: 86.00
				Recall: 17.19
			Corel-1500	Precision: 83.20
				Recall: 16.64
			Caltech-256	Precision: 38.98
				Recall: 7.796
2019	Modified color difference histogram [79]	Euclidean distance	Wang subset	Precision:75.33
				Recall: 18.61
				Bull’s eyes Percentage:48.74
2019	A fusion of HSV color space, GLCM, LBP, and normalized moment inertia [80]	Euclidean distance	Corel-1k	Accuracy:69.7
				Recall: 69.1
				Fmeasure: 69.4
			AT&T face dataset	Accuracy: 74.6
				Recall: 70.9
				Fmeasure: 72.7
			FD-XJ face dataset	Accuracy: 62.9
				Recall: 61.8
				Fmeasure: 61.8
2020	Intensity variation descriptor by fusion of HSV and RGB color features, edges, and intensity variations-based texture features [1]	Extended L1 distance	Corel-5K	Precision: 66.9
				Recall: 8.03
				Fmeasure: 14.34
			Corel-10K	Precision: 56.88
				Recall: 6.83
				Fmeasure: 12.20
2017	Fusion low order color moments, DCT-based texture features, and salient region-based shape features with the SVM classifier [81]	Normalized matching ratio	Corel-1000	Precision 78.1
				Recall 17.2
			Oxford flowers	Precision 82.3
				Recall 18.7
			Caltech-256	Precision 47.0
				Recall 20.3
2017	A fusion of color moments, HSV histogram, co-occurrence matrix, wavelet moments, and chain code features [82]	Manhattan distance	Wang	Precision: 82.4