You are on page 1of 4

Multimedia Image Retrieval System by Combining CNN with

Handcraft Features in Three Different Similarity Measures

Abstract
With the fast improvement of PC innovation and the expanding of
multimedia information on the Internet, how quickly find the desired
information in the massive data becomes a hot issue. Image retrieval can
be used to retrieve similar images, and the effect of image retrieval
depends on the selection of image features to a certain extent. From
inquiry by an image to retrieve of pertinent images, CBIR has diverse
stages. In any case, features extraction from the images is one of the
significant stages. As of late, CNN indicates great outcomes in the PC
vision domain because of the ability of feature extraction from the
images. We propose WNAHVF a combined Weighted and Normalized
AlexNet with Handcrafted Visual Features (BOF and LNP) for extract
feature from the image and use those vectors for image retrieval and
classification. We test WNAHVF method on two general datasets (Corel-
1k and Corel-10k) and one medical dataset. Our outcomes demonstrate
combining BOF and LNP with AlexNet enhances the accuracy and give
better results in general and medical image datasets in retrieval and
classification problems. This algorithm gives results which are superior to
existing strategies and give an extraordinary accomplishment in the
image retrieval and classification domains.
EXISTING SYSTEM

Query Image Pre-Wavelet


processing
Decompose DCT

Nearest
Neighbourhood
Detection

Similarity
Mapping

Similar Image Comparision Extracting


Retrive Features

Fig 1. Existing System

DISADVANTAGES OF EXISTING SYSTEM:

 Existing method achieved low performance


 Prediction efficiency low

PROPOSED SYSTEM

Pre-Processing Wavelet K-Means


Query Image
Decompose Clustering

Feature
Extraction

Similar Image Feature


Recalling Matching

Fig 2. Proposed System


The Proposed system extracts the feature vectors from the images
in the database and stores the feature vectors in the database. In the online
stage, the user inputs an image as inquiry image, and afterward, the
algorithm extricates the visual component vector from this image and
measure the similarity distance between the element vector of inquiry
image and all element (feature) vectors in the database which extract in
the off-line stage. At that point, we sort the distance in a rising manner
and retrieve top k images. For feature extraction, we proposed WNAHVF
algorithm and for similarity distance, we use L1, Euclidean, and
Cityblock.

ADVANTAGES OF PROPOSED SYSTEM:

 Comparatively more accurate


 Quality of the Image is increasing
 High Performance

SOFTWARE REQUIREMENTS:

 Operating system : Windows XP/7


 Coding Language : MATLAB
 Tool : MATLAB R2013A

REFRENCES:
1. T. He, Y. Wei, Z. Liu, G. Qing and D. Zhang. (2018): Content Based
Image Retrieval Method Based on SIFT Feature. ICITBS, Xiamen China,
pp. 649-652.
2. Shah, R. Naseem, Sadia, S. Iqbal and M. Shah. (2017): Improving
CBIR accuracy using convolutional neural network for feature extraction.
13th International Conference on Emerging Technologies (ICET).
3. M. Alkhawlani, M. Elmogy and H. Elbakry. (2015): Content-Based
Image Retrieval using Local Features Descriptors and Bag-ofVisual
Words. International Journal of Advanced Computer Science and
Applications, vol. 6, no. 9.
4. A. Krizhevsky, I. Sutskever and G. Hinton. (2017): ImageNet
classification with deep convolutional neural networks. Communications
of the ACM, vol. 60, no. 6, pp. 84-90.
5. A. Alzu’bi, A. Amira, and N. Ramzan. (2015): Semantic content-based
image retrieval: A comprehensive study. Journal of Visual
Communication and Image Representation, vol. 32, no. July, pp. 20–54.

You might also like