Call For Paper Volume:4 Issue:9 Sep'2017 |

Truncated DCT Image BTC for Image Retrieval

Publication Date : 06/12/2015



Author(s) :

Dr.Ravindra patil , NST SAI , Shailesh Sangle.


Volume/Issue :
Volume 2
,
Issue 12
(12 - 2015)



Abstract :

Histogram based color features vector size is large which require more memory to save it. To make a feature vector compact and effective in terms of retrieval performance Block Truncation Coding (BTC) for the color images are proposed by the researchers. To calculate the mean color value of each color plane whole pixels of the color planes are used which is also time consuming. Time complexity to calculate the mean color value of color planes are reduced by using the smaller size of image.In this paper we discuss proposed two techniques of Content Based Image Retrieval (CBIR) to improve color feature using Truncated Discrete Cosine Transform (DCT) image Block Truncation Coding . We discuss the comparative study of full image BTC, truncated DCT image BTC and truncated DCT image sub block BTC using RGB, YCbCr, YUV, CXY, CIEL*a*b*, R’G’I color planes. Transformed DCT image low frequency coefficients are used to generate the truncated DCT image. In this paper we take up to fifth level DCT truncation and each level truncated DCT image BTC performance is compared with the full image BTC. Proposed techniques are tested over the image database which includes 1200 images having 15 classes. Similarity between query image feature vector and database feature vector measured here using Euclidean distance and Bray Curtis distance similarity. Performance of the proposed approaches measured by using overall average precision and recall cross over point


No. of Downloads :

5


Indexing

Web Design MymensinghPremium WordPress ThemesWeb Development

Truncated DCT Image BTC for Image Retrieval

December 2, 2015