Image Hashing for DWT SPIHT Coded Images

1
Image Hashing for DWT SPIHT Coded Images

• ???

2
Outline

• Image hashing
• The significance maps from SPIHT
• The SPIHT-autocorrelogram
• Distance(similarity) measure
• Experimental results
• Future work

3
Image hashing

• Watermarking
• Content-based image retrieval(CBIR)
• Image hashing

4
The significance maps from SPIHT

• SPIHT
• Initialization
• Sorting pass
• Refinement pass
• Quantization-step update
• output bit stream

5
The significance maps from SPIHT

• In sorting pass, we can get the significance of
  each entry in LIP and LIS(A type and B type). So
  we form the significance maps according to the
  above property.
• Only the last 4 subbands are considered

6
The significance maps from SPIHT

• examples

7
The significance maps from SPIHT

• example

LIP
LIS(A)
LIS(B)
8
The SPIHT-autocorrelogram

• Histogram-based method in CBIR
• ex CCV,color correlogram,etc
• property contain both color and spatial
  information
• resistant to geometric distortion

9
The SPIHT-autocorrelogram

• Count the autocorrelogram of 1s for each
  significance map
• let a significance map M be a mxm matrix
• , means its value

10
The SPIHT-autocorrelogram

• Count the autocorrelogram of 1s for each
  significance map
• let a distance
• the autocorrelogram of 1s of M is defined as

11
The SPIHT-autocorrelogram

• example

12
Distance(similarity) measure

• For the significance maps or the
  SPIHT-autocorrelograms, convert them to an
  one-dimension vector as our hash.

13
Distance(similarity) measure

• Distance measure
• using L1 distance
• let H and H be the hashes of two iamges
• Hi means the value of the ith entry in H
• the L1 distance between two hashes is defined
  as

14
Experimental Results

• Setup
• database 900images(100 different images and 800
  
• attacked images)
• color space YCbCr
• DWT 9/7f
• level 5
• the thresholds the first 3 thresholds
• sign maps per image 3343108

15
Experimental Results
 

• Attack modes

16
Experimental Results

• Example of attacked images

17
Experimental Results

• Performance measure
• The efficiency of retrieval proposed by
• Kankanhalli
• N the number of ground truth
• T the first T similar image we consider in
  retrieval
• n the number of matched images in retrieval

18
Experimental Results

• Results
• the performance between significance maps and
  SPIHT-autocorrelogram

19
Experimental Results

• Results an example query by 0.jpg

20
Future work

• More attack modes
• Reading more papers
• Comparing with papers