TY - GEN
T1 - Content based image retrieval by IPP algorithm
AU - Song, Jia
AU - He, Zunwen
N1 - Publisher Copyright:
© 2015 IEEE.
PY - 2015/7/14
Y1 - 2015/7/14
N2 - In order to realize the content-based image retrieval (CBIR), some characteristics of the images should be extracted like color, texture and shape. The extremely important thing in CBIR is to search the most similar database images to match the query image, which needs to improve the precision. This paper proposes an Improving Precision Priority (IPP) algorithm integrating vital features and the query method to improve performance. Proposed IPP algorithm has two phases. In the first phase, both of the query image and database images are divided into several blocks respectively. After that, we calculate the color histogram of each block. Then we take Euclidean distance to compare the similarities to complete the first round of retrieval. To calculate the distance, we allocate different blocks to different weights, the blocks of the central part always containing much useful information should be allocated more weight. And the surrounding part are allocated less and the corners have the smallest weight. All of the distances of the small blocks are accumulated together to be the distance of the whole image. In this phase we can retrieve some related images from the database denoting as result A. In the second phase, shape characteristics of result A are extracted using Hu moment invariants. After that, we calculate the invariant moments similarities between the query image and those of result A images. The most similar images are shown as the final result. IPP algorithm can increase the precision.
AB - In order to realize the content-based image retrieval (CBIR), some characteristics of the images should be extracted like color, texture and shape. The extremely important thing in CBIR is to search the most similar database images to match the query image, which needs to improve the precision. This paper proposes an Improving Precision Priority (IPP) algorithm integrating vital features and the query method to improve performance. Proposed IPP algorithm has two phases. In the first phase, both of the query image and database images are divided into several blocks respectively. After that, we calculate the color histogram of each block. Then we take Euclidean distance to compare the similarities to complete the first round of retrieval. To calculate the distance, we allocate different blocks to different weights, the blocks of the central part always containing much useful information should be allocated more weight. And the surrounding part are allocated less and the corners have the smallest weight. All of the distances of the small blocks are accumulated together to be the distance of the whole image. In this phase we can retrieve some related images from the database denoting as result A. In the second phase, shape characteristics of result A are extracted using Hu moment invariants. After that, we calculate the invariant moments similarities between the query image and those of result A images. The most similar images are shown as the final result. IPP algorithm can increase the precision.
KW - IPP algorithm
KW - block color histogram
KW - image retrieval
KW - moment invariant
KW - shape
UR - http://www.scopus.com/inward/record.url?scp=84943183249&partnerID=8YFLogxK
U2 - 10.1109/CIVEMSA.2015.7158604
DO - 10.1109/CIVEMSA.2015.7158604
M3 - Conference contribution
AN - SCOPUS:84943183249
T3 - 2015 IEEE International Conference on Computational Intelligence and Virtual Environments for Measurement Systems and Applications, CIVEMSA 2015
BT - 2015 IEEE International Conference on Computational Intelligence and Virtual Environments for Measurement Systems and Applications, CIVEMSA 2015
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2015 IEEE International Conference on Computational Intelligence and Virtual Environments for Measurement Systems and Applications, CIVEMSA 2015
Y2 - 12 June 2015 through 14 June 2015
ER -