[Retracted] Object-Based Image Retrieval Using the U-Net-Based Neural Network

Table 1

Comparison of the existing image retrieval techniques.

S. no.	Author	Year	Method/methodology	Comments
1	Liying and Lirong [13]	2017	Scale-invariant feature transform (SIFT)	Performs retrieval on more than a single query to increase the retrieval accuracy.
			Convolutional neural network (CNN)
2	Dhotre et al. [4]	2017	The color feature extracted through CHRIR and wavelet transform performed using multilevel Haar wavelet transform (MHWT)	Proved to be a faster retrieval method on an image database with one of the physical features. Works more accurately with increased retrieval speed and minimized time.
3	Jayanthi and Karthikeyan [2]	2015	FCTH, CEDD, HWT, and DWT using fuzzy linking and Gabor filters	Database with 1000 color images results in better recall and average precision of retrieval.
4	Thepade and Shinde [3]	2015	Haar wavelet transforms with Canny edge detection based on shape features using gradient techniques such as Prewitt, Laplace, and Sobel, and the slope magnitude technique with the Manhattan similarity function	Database with 350 color images
				Frei-Chen and Sobel give better performance than the other algorithms that used Canny implementation.
5	Jayanthi and Karthikeyan [2]	2015	HWT and DWT using Gabor filters and fuzzy linking	Gives good results in average precision and recall value.
6	Gupta and Kushwah [8]	2015	Haar discrete wavelet transform (H-DWT), gray level co-occurrence matrix (GLCM)	Improved results in comparison with previous methods.
			Support vector machine (SVM)
7	Agarwal et al. [5]	2014	Used color edge detection and DWT	Database used was Wang’s image database. Gave high precision and recall values.
8	Agarwal et al. [5]	2014	Color edge detection, DWT	High precision and recall indicate an exemplary retrieval system
			Canny edge detection
9	Ying Chen and Wu [6]	2011	Uses optical flow to extract the information from the video; extraction process with Haar wavelet	Locates the feature almost near the query point using index values.
10	Chatzichristofis et al. [7]	2010	The method based on color and edge directivity descriptor (CEDD)	Demonstrated successful retrieval on benchmark datasets.
			Utilizes the binary Haar wavelet transform for extraction
11	Quellec et al. [12]	2010	The multidimensional wavelet filter bank	Can be used in a different dimension of signal and different lattices.
12	Verma et al. [1]	2009	Texture analysis-based scheme	Own dataset with 100 color images of size 256 × 256 pixels each. Accuracy up to 73%.
			I level Haar wavelet used for image decomposition F-norm theory to decrease the dimension of the extracted feature; fuzzy logic similarity measure used
13	Huang et al. [10]	2005	Lifting scheme F-norm theory	Good at retrieval.
14	Wong et al. [9]	2005	AdaBoost-based face defection method and the lifting wavelet transform (LFWT) technique	Efficient with small memory to detect the face. Suits best for multimedia applications
15	Munjal and Bhatia [17]	2019	UCID dataset used, CEDD and Gabor wavelet transform (GWT) for feature vector	Accuracy = 91.9%.
16	Varish et al. [18]	2020	Fusion of histograms of gradients and invariant moments, Corel 1K and GHIM-10K dataset used for validation	Precision = 89% and 90%.
				Recall = 17.80% and 3.60.
				F-score = 29.49% and 6.92%.
17	Wadhera and Agarwal [19]	2020	3D center symmetric LBP + Gaussian filter + gray level co-occurrence matrix (GLCM), STex texture, and ESSEX face database	Precision = 61% and 97.4%.
				Retrieval rate = 97.4%.
18	Ajam et al. [20]	2019	LBP + HSV + entropy, Corel 10K and Corel 5K databases used	Retrieval rate = 59.51% and 49.13%.
19	Xiaobo et al. [21]	2021	Adaptive threshold + directional LBP, and Corel 1K database	Precision = 67.63%.