fractal video compression 碎形視訊壓縮方法 chia-yuan chang 張嘉元 department of applied...

30
Fractal Video Compression 碎碎碎碎碎碎碎碎 Chia-Yuan Chang 張張張 Department of Applied Mathematics National Sun Yat-Sen University Kaohsiung, Taiwan

Upload: marilyn-ellis

Post on 05-Jan-2016

232 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Fractal Video Compression 碎形視訊壓縮方法 Chia-Yuan Chang 張嘉元 Department of Applied Mathematics National Sun Yat-Sen University Kaohsiung, Taiwan

Fractal Video Compression碎形視訊壓縮方法

Chia-Yuan Chang

張嘉元Department of Applied Mathematics

National Sun Yat-Sen University

Kaohsiung, Taiwan

Page 2: Fractal Video Compression 碎形視訊壓縮方法 Chia-Yuan Chang 張嘉元 Department of Applied Mathematics National Sun Yat-Sen University Kaohsiung, Taiwan

Topics

• Introduction

• Our approach

• Simulation

• Conclusions

Page 3: Fractal Video Compression 碎形視訊壓縮方法 Chia-Yuan Chang 張嘉元 Department of Applied Mathematics National Sun Yat-Sen University Kaohsiung, Taiwan

INTRODUCTION

• Standardization of algorithm -- MPEG

• Quad-tree structure

• Slicing floorplan tree

• Fractal dimension

Page 4: Fractal Video Compression 碎形視訊壓縮方法 Chia-Yuan Chang 張嘉元 Department of Applied Mathematics National Sun Yat-Sen University Kaohsiung, Taiwan

Standardization of algorithm• MPEG

– video layers

• I-picture: Intraframe, JPEG DCT, lower compression ratio

• P-picture: Predicted frame, motion compensation

• B-picture: Bi-directional frame,higher compression ratio

Page 5: Fractal Video Compression 碎形視訊壓縮方法 Chia-Yuan Chang 張嘉元 Department of Applied Mathematics National Sun Yat-Sen University Kaohsiung, Taiwan

– display order

I B B B B B B B B BP P P

group of pictures

time

Page 6: Fractal Video Compression 碎形視訊壓縮方法 Chia-Yuan Chang 張嘉元 Department of Applied Mathematics National Sun Yat-Sen University Kaohsiung, Taiwan

– coding order

I P B B P B B P B BB B B

group of pictures

time

Page 7: Fractal Video Compression 碎形視訊壓縮方法 Chia-Yuan Chang 張嘉元 Department of Applied Mathematics National Sun Yat-Sen University Kaohsiung, Taiwan

– motion compensationtime

Forward predictionreference picture

Current pictureBackward prediction

reference picture

Page 8: Fractal Video Compression 碎形視訊壓縮方法 Chia-Yuan Chang 張嘉元 Department of Applied Mathematics National Sun Yat-Sen University Kaohsiung, Taiwan

– disadvantages

• buffer and time control

• encoding: the fixed block size

• DCT: filter high frequency (like edge)

Page 9: Fractal Video Compression 碎形視訊壓縮方法 Chia-Yuan Chang 張嘉元 Department of Applied Mathematics National Sun Yat-Sen University Kaohsiung, Taiwan

Quad-tree structure

• basic definition

– top-down : segment

– bottom-up : merge

• application

– Vector Quantization (VQ)

• disadvantage

– efficiency

Page 10: Fractal Video Compression 碎形視訊壓縮方法 Chia-Yuan Chang 張嘉元 Department of Applied Mathematics National Sun Yat-Sen University Kaohsiung, Taiwan

Slicing floorplan tree

• The Recursive Split Algorithm– Start with R containing a single rectangular

patch that covers the entire frame– Repeat n-1 times Step 1), 2), 3)– 1) Search R for the rectangle r with the largest

error er, and remove it from R.

– 2) Split r into two rectangles r1, r2 such that er1 + er2 is minimized.

– 3) Add r1, r2 to R.

Page 11: Fractal Video Compression 碎形視訊壓縮方法 Chia-Yuan Chang 張嘉元 Department of Applied Mathematics National Sun Yat-Sen University Kaohsiung, Taiwan

• disadvantage

– each two-frames has own mask

– noise effect

Page 12: Fractal Video Compression 碎形視訊壓縮方法 Chia-Yuan Chang 張嘉元 Department of Applied Mathematics National Sun Yat-Sen University Kaohsiung, Taiwan

Fractal dimension

• Introduction– estimate length of coastline

– general formula

– the measurement, analysis, and classification of shape and texture

DFL 1)(

1rDNr or DNr

r

log

log1

Page 13: Fractal Video Compression 碎形視訊壓縮方法 Chia-Yuan Chang 張嘉元 Department of Applied Mathematics National Sun Yat-Sen University Kaohsiung, Taiwan

• Box counting approach (3-D space)– image size : M x M– box size : s x s– ratio : r = s / M– box number in ( i, j) grid

– total box number

– FD equation r

rND1log

log

1, kljinr

jinNji

rr ,,

Page 14: Fractal Video Compression 碎形視訊壓縮方法 Chia-Yuan Chang 張嘉元 Department of Applied Mathematics National Sun Yat-Sen University Kaohsiung, Taiwan
Page 15: Fractal Video Compression 碎形視訊壓縮方法 Chia-Yuan Chang 張嘉元 Department of Applied Mathematics National Sun Yat-Sen University Kaohsiung, Taiwan

Our approach

• Fractal Dimension Estimation

• Slicing Floorplan Segmentation

• Compression

• Decompression

Page 16: Fractal Video Compression 碎形視訊壓縮方法 Chia-Yuan Chang 張嘉元 Department of Applied Mathematics National Sun Yat-Sen University Kaohsiung, Taiwan

Mask processing

m+ 1sucessive

images

mdifference

images

m imagesamples

featuremap

F(i, j)

Gray-levelscale

Masksegmentation

compute fractal dimension

Page 17: Fractal Video Compression 碎形視訊壓縮方法 Chia-Yuan Chang 張嘉元 Department of Applied Mathematics National Sun Yat-Sen University Kaohsiung, Taiwan

• Fractal dimension estimation– Basic definition

• image sequence: I(x, y, t),

• a group image frames: F1, F2, …, Fm+1,

• reference frame : Fr

• frame difference : Diff(x, y, t)• difference volume : V• voxel (x, y, t)• feature map : F(x, y)

Ntyx ,,

Page 18: Fractal Video Compression 碎形視訊壓縮方法 Chia-Yuan Chang 張嘉元 Department of Applied Mathematics National Sun Yat-Sen University Kaohsiung, Taiwan

– A modified box-counting approach• window volume size : mxmxm• cubes size : axaxa. • scaling factor s,• the fractal dimension for the voxel (x, y, t)

mas

1,, ijtyxcont

2,,12

),,(mrqpm

rtqypxcontC

s

CFD tyx 1log

log),,(

Page 19: Fractal Video Compression 碎形視訊壓縮方法 Chia-Yuan Chang 張嘉元 Department of Applied Mathematics National Sun Yat-Sen University Kaohsiung, Taiwan

¡´ (x ,y ,m /2 )

Diff (x, y ,1)

Diff (x ,y ,m )

Diff (x, y ,m /2)

¡´ (x, y, m )

n

nm

m

aa

Difference volumeV

a

¡´ (x ,y ,1)

¡´ ¡´

¡´

¡´ ¡´

¡´

T h e r e l a t i o n s h i p b e t w e e n t h e f r a m e d i m e n s i o n n , f r a m e d i f f e r e n c e

D i f f ( x , y , t ) , t h e n u m b e r o f f r a m e d i f f e r e n c e in a m a c r o b l o c k m , a n d

t h e d i m e n s i o n o f t h e m e a s u r i n g c u b e a .

Page 20: Fractal Video Compression 碎形視訊壓縮方法 Chia-Yuan Chang 張嘉元 Department of Applied Mathematics National Sun Yat-Sen University Kaohsiung, Taiwan

• Slicing floorplan segmentation.– Start with R containing a single rectangular patch

that covers feature map F(i, j).• 1) search R for the rectangle r with the largest

variance Vr if Vr < Vt then go to Step 4 else remove it from R.

• 2) split r into two rectangles r1, r2 such that is maximized

• 3) add r1, r2 to R, and go to Step 1

• 4) check the mean value of each block. If Mr > Mt then segment Mr to smaller blocks else exit.

M rrM21

Page 21: Fractal Video Compression 碎形視訊壓縮方法 Chia-Yuan Chang 張嘉元 Department of Applied Mathematics National Sun Yat-Sen University Kaohsiung, Taiwan

Motion estimation

referenceimage

predicted image usingreference image

xi

yi

Page 22: Fractal Video Compression 碎形視訊壓縮方法 Chia-Yuan Chang 張嘉元 Department of Applied Mathematics National Sun Yat-Sen University Kaohsiung, Taiwan

Compression

referenceimage Fr

Mask

motionestimation

motionvector of

each blockin mask

image Fi

Page 23: Fractal Video Compression 碎形視訊壓縮方法 Chia-Yuan Chang 張嘉元 Department of Applied Mathematics National Sun Yat-Sen University Kaohsiung, Taiwan

Decompression

referenceimage Fr

Mask

motion vector ofeach block in the

mask

image Fi

Page 24: Fractal Video Compression 碎形視訊壓縮方法 Chia-Yuan Chang 張嘉元 Department of Applied Mathematics National Sun Yat-Sen University Kaohsiung, Taiwan

Simulation

• test image sequence

– Claire

– football

– Noisy Claire (25db Gaussian noises)

– Noisy football (20db Gaussian noises)

• comparison

– MPEG

Page 25: Fractal Video Compression 碎形視訊壓縮方法 Chia-Yuan Chang 張嘉元 Department of Applied Mathematics National Sun Yat-Sen University Kaohsiung, Taiwan

(a) (b)

(c) (d)

Fig. 2 (a) The 1st frame in the Claire sequence, (b) The 1st frame in the football sequence, (c) 25 dB Gaussian noises are added to part (a), (d) 20 dB Gaussian noises are added to part (b).

Page 26: Fractal Video Compression 碎形視訊壓縮方法 Chia-Yuan Chang 張嘉元 Department of Applied Mathematics National Sun Yat-Sen University Kaohsiung, Taiwan

Fig. 3 (a) – (d) The feature map representing the fractal dimension for the image sequences in Fig. 2 (a) – (d), respectively, after normalizing to 255.

(a) (b)

(c) (d)

Page 27: Fractal Video Compression 碎形視訊壓縮方法 Chia-Yuan Chang 張嘉元 Department of Applied Mathematics National Sun Yat-Sen University Kaohsiung, Taiwan

Fig. 4 (a) – (d) The slicing floorplan segmentation maskcorresponding to the image sequences in Fig. 2 (a) – (d), respectively.

(a) (b)

(c) (d)

Page 28: Fractal Video Compression 碎形視訊壓縮方法 Chia-Yuan Chang 張嘉元 Department of Applied Mathematics National Sun Yat-Sen University Kaohsiung, Taiwan

Image sequences Total block

numbers

Bit rate

(bit/pixel)

Average PSNR

(dB)

Claire 137

199

325

.044526

.052443

.068533

35.71311

35.88222

36.31750

Noisy Claire 88

142

233

.040900

.046518

.058139

27.94569

28.02084

28.13815

Football 122

335

572

.044816

.072016

.102280

21.82529

23.26751

24.22758

Noisy Football 16

45

98

.062575

.066278

.073046

16.30597

17.11834

17.80838

Table 1. The relationship between the number of blocks, the compression ratio, and

PSNR for the image sequence.

Page 29: Fractal Video Compression 碎形視訊壓縮方法 Chia-Yuan Chang 張嘉元 Department of Applied Mathematics National Sun Yat-Sen University Kaohsiung, Taiwan

Image sequences Compression

methodology

Bit rate

(bit/pixel)

Average PSNR

(dB)

Claire MPEG

Our approach

.143985

.044526

35.38568

35.71311

Noisy Claire MPEG

Our approach

.143985

.040900

23.34401

27.94569

Football MPEG

Our approach

.143985

.035877

19.25690

19.26762

Noisy Football MPEG

Our approach

.143985

.062575

16.32646

16.30597

Table 2. Comparison of PSNR and bit rate for our approach and MPEG

Page 30: Fractal Video Compression 碎形視訊壓縮方法 Chia-Yuan Chang 張嘉元 Department of Applied Mathematics National Sun Yat-Sen University Kaohsiung, Taiwan

Conclusions

• Our algorithm can get higher compression ratio than MPEG in the same average PSNR for the same image sequence.

• Future research

– compression speed improvement