IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 7, NO. 1, FEBRUARY 2005, pp. 105-113

22
Image Retrieval Tool Using Automatic Image Segmentation and Stepwise Boolean AND Matching IEEE TRANSACTIONS ON MULTIMEDIA, V OL. 7, NO. 1, FEBRUARY 2005, pp. 1 05-113 ByoungChul Ko and Hyeran Byun Reporter: Jen-Bang Feng

description

FRIP: A Region-Based Image Retrieval Tool Using Automatic Image Segmentation and Stepwise Boolean AND Matching. IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 7, NO. 1, FEBRUARY 2005, pp. 105-113 ByoungChul Ko and Hyeran Byun Reporter: Jen-Bang Feng. Outline. Image Retrieval - PowerPoint PPT Presentation

Transcript of IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 7, NO. 1, FEBRUARY 2005, pp. 105-113

Page 1: IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 7, NO. 1, FEBRUARY 2005, pp. 105-113

FRIP: A Region-Based Image Retrieval Tool Using Automatic Image Segmentation and Stepwise Boolean AND Matching

IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 7, NO. 1, FEBRUARY 2005, pp. 105-113ByoungChul Ko and Hyeran ByunReporter: Jen-Bang Feng

Page 2: IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 7, NO. 1, FEBRUARY 2005, pp. 105-113

2

Outline

Image Retrieval Content-Based Image Retrieval The Proposed Scheme Experimental Results

Page 3: IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 7, NO. 1, FEBRUARY 2005, pp. 105-113

3

Image Retrieval

ImageDB

Image retrievalscheme FeaturesFeaturesFeaturesFeaturesFeaturesFeaturesFeatures

QueryImage

Image retrievalscheme

Feature

Compare SearchingResults

Page 4: IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 7, NO. 1, FEBRUARY 2005, pp. 105-113

4

Content-Based Image Retrieval From text-based retrieval scheme

WWW search engine Query-by-image in early 90’s From global to local (region)

Region Of Interest

Page 5: IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 7, NO. 1, FEBRUARY 2005, pp. 105-113

5

The Proposed Scheme

1. Image Segmentation Two-Level Segmentation Using Adaptive Circular Filter a

nd Bayes’ Theorem Iterative Level Using Region Labeling and Iterative Regio

n Merging2. Feature extraction

Color Texture Normalized Area Shape and Location

3. Stepwise Similarity Matching

Page 6: IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 7, NO. 1, FEBRUARY 2005, pp. 105-113

6

Two-Level Segmentation Using Adaptive Circular Filter and Bayes’ Theorem

Adaptive Circular Filter

Image(RGB)

Image(CIE Lab)

SmoothedImage(CIE Lab)

Remove middle frequency

Color histogram

Separate regions by circular filters

RegionsRegionsRegionsRegions

Page 7: IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 7, NO. 1, FEBRUARY 2005, pp. 105-113

7

Two-Level Segmentation Using Adaptive Circular Filter and Bayes’ Theorem

a is similar to c in colorbut a is closer to b than c Example of circular filtering process

Page 8: IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 7, NO. 1, FEBRUARY 2005, pp. 105-113

8

Two-Level Segmentation Using Adaptive Circular Filter and Bayes’ Theorem

xyyx

CyxyxM

MyxMMyxM

MyxMyxM

cc

MccCP

CcPCPCcPCP

CcPCPcCP

,

,,

,,

,,

else

then ,5.0| if

||

||

Three circular filters3x3, 7x7, 11x11

CM: the most frequently observed histogram binsCM: other binscx,y: center value of CM

MC: the major class color

Page 9: IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 7, NO. 1, FEBRUARY 2005, pp. 105-113

9

Two-Level Segmentation Using Adaptive Circular Filter and Bayes’ Theorem

division according to the edge distribution Selected filter, 3x3, 7x7, 11x11

Segmentation result Final segmented image

Page 10: IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 7, NO. 1, FEBRUARY 2005, pp. 105-113

10

Iterative Level Using Region Labeling and Iterative Region Merging

Image(RGB)

Image(CIE Lab)

SmoothedImage(CIE Lab)

Remove middle frequency

Color histogram

Separate regions by circular filters

RegionsRegionsRegionsRegions

RegionsRegions

Merge regions

Page 11: IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 7, NO. 1, FEBRUARY 2005, pp. 105-113

11

Iterative Level Using Region Labeling and Iterative Region Merging

N

i

imbb

imaa

imLL TRRRRRRIf

For the N neighbor regions

Then merge the regions

If the number of regions is larger than 30Then increase the threshold and repeat the circular filter

Page 12: IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 7, NO. 1, FEBRUARY 2005, pp. 105-113

12

Feature extraction Color

Average AL, Aa, Ab

Variance VL, Va, Vb

Color distance of Q and T

2,,,,

,

VbVaVLCCC

AbAaAlCCC

CTQ TQTQd

Page 13: IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 7, NO. 1, FEBRUARY 2005, pp. 105-113

13

Feature extraction Texture

Biorthogonal wavelet frame (BWF) The X-Y directional amplitude Xd, Yd

The distance in texture

T

T

Q

QTTQ Xd

Yd

Xd

Ydd ,

Page 14: IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 7, NO. 1, FEBRUARY 2005, pp. 105-113

14

Feature extraction Normalized Area

NPQ =

(Size of the region) / (Size of the image)

TQNArea

TQ NPNPd ,

Page 15: IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 7, NO. 1, FEBRUARY 2005, pp. 105-113

15

Feature extraction Shape and Location

The global geometric shape feature eccentricity

Estimate the bounding rectangle for each segmented region

For the major axis Rmax and minor axis Rmin

max

min

max

min

T

T

Q

Q

R

R

R

RE

Page 16: IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 7, NO. 1, FEBRUARY 2005, pp. 105-113

16

Feature extraction Shape and Location

The local geometric shape feature MRS (modified radius-based shape signature)

invariant under shape’s scaling, rotation, and translation

Page 17: IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 7, NO. 1, FEBRUARY 2005, pp. 105-113

17

Feature extraction Shape and Location

The local geometric shape feature MRS (modified radius-based shape signature)

Extracts 12 radius distance values

2

1 1

,

11

,min

N

i i

iN

j j

jC

ckwisecountercloclockwiseMRS

TQ

T

Q

T

Q

NNd

ddd

Page 18: IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 7, NO. 1, FEBRUARY 2005, pp. 105-113

18

Stepwise Similarity Matching

r

i

p

j

ij

ijjj tqDwYXSim

1 1

,,

Page 19: IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 7, NO. 1, FEBRUARY 2005, pp. 105-113

19

Experimental Results

query: flower best case

Page 20: IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 7, NO. 1, FEBRUARY 2005, pp. 105-113

20

Experimental Results

query: shipworst case

Page 21: IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 7, NO. 1, FEBRUARY 2005, pp. 105-113

21

Page 22: IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 7, NO. 1, FEBRUARY 2005, pp. 105-113

22