This document presents an effective image retrieval system using dot-diffused block truncation coding (DDBTC) features. It discusses two main methods of image retrieval: text-based and content-based. Content-based retrieval analyzes images based on color, texture, or form but requires long processing times. The document then describes DDBTC, which divides color images into blocks, computes quantizers for each block, and generates a bitmap image. This approach reduces processing time for image encoding and retrieval compared to other methods. Experimental results showed the DDBTC features provided an effective image retrieval system for large image databases.
2. Introduction
Computer system for browsing, searching and
retrieving images from a large database of digital
images
Two methods
Text Based Image Retrieval
Content Based Image Retrieval
TBIR
Requires text as input to search for images
Uses keywords such as image name, date etc
3. Contd..
CBIR
Input as query image
Search similar images by colour, texture or form as
a comparison
Requires long processing time due to amount of
images to be analyzed in database
4. Motivation
Due to the increase of online users on the Internet,
the amount of collections of digital images have
grown continuously during this period. So, it is
necessary to develop an appropriate system to
manage effectively these collections.
5. Block Truncation Coding
Technique for image compression
Divide image into multiple non-overlapped image
block
Each block represented by two quantizers
High & low mean value
Bitmap image
Perform thresholding operation using mean value
to generate bitmap image
6. BTC Image Coding Using Vector
Quantization
Source encoding the output of a BTC
Overhead statistical information
Truncated block
If vector quantization used on either one of
output- bit rate reduces upto 1.5 bits/pel
Vector quantization used on both output- bit rate
reduces upto 1.0 bits/pel
8. BTC-VQ-DCT Hybrid Coding of
Digital Images
Simple computation & edge preservation of BTC
High fidelity & high compression ratio of adaptive
DCT
High compression ratio & good subjective
performance of VQ
Advantage
Remove both inter block and intra block redundancy
Computational complexity is less than DCT or VQ
Compression ratio of 10:1
9. Absolute Moment Block Truncation
Coding
Digitized image divided into block
Quantized each block results in same sample mean &
same sample first absolute central moment
Mean-information about central tendancy
First absolute central moment- information about
dispersion from mean
Advantage
Processing time is reduced at transmitter and receiver-
fast BCT
11. Contd..
Color image of size MxN in RGB divided into
several non-overlapping image blocks
Each image block processed independently
DDBTC encoder generates two extreme
quantizers
Minimum quantizers
•Maximum quantizers
13. Contd..
Diffused weight
•Bitmap image obtained by
•Decoder receives set of min and max quantizer and
bitmap
•Replace 0 with minimum quantizer value and 1 with
maximum quantizer
14. References
[1] J. Guo, H. Prasetyo, and N. Wang, “Effective Image
Retrieval System Using Dot-Diffused Block Truncation
Coding Features”, IEEE Trans. multimedia, vol. 17, no. 9,
september 2015
[2] V. R. Udpikar and J. P. Raina, “BTC image coding using
vector quantization,” IEEE Trans Commun., vol. COM-35,
no. 3, pp. 352–256, Sep.1987.
[3] Y. Wu and D. Coll, “BTC-VQ-DCT hybrid coding of digital
images,” IEEE Trans. Commun., vol. 39, no. 9, pp. 1283–
1287, Sep. 1991.
[4] M. D. Lema and O. R. Mitchell, “Absolute moment block
truncation coding and its application to color images,” IEEE
Trans. Commun., vol. COM-32, no. 10, pp. 1148–1157,
Oct. 1984.