sample thesis defense ppt
-
Upload
christian-arcelo -
Category
Documents
-
view
262 -
download
0
Transcript of sample thesis defense ppt
-
7/24/2019 sample thesis defense ppt
1/26
Low Resolution Image
Enhancement
Aquino, Juan Miguel A.
Arcelo, Christian
-
7/24/2019 sample thesis defense ppt
2/26
Objective of the study
Main objective: enhance still images taken from standard lo resolution
la!to! cameras.
"ub objectives:
#esearch and find noise filtering technique hich can be used to com!are andsubstitute to $uassian %limination
&ind and a!!ly an image illumination correction method hich can be used to
com!are and substitute to Multi'"cale #etine( Algorithm
find and a!!ly best !arameter to reconstruct the lo resolution image
find a method hich can be used to visually and mathematically com!are the lo
resolution image to the !rocessed image
-
7/24/2019 sample thesis defense ppt
3/26
"co!e and )imitations
*+(+- lo resolution la!to! camera
+- unique lo resolution images as taken.
"tanding or sitting !osition and facing the camera, smiling or neutral
emotion. "ubjects face must be visible ithout obstruction such as
hat, scarf, hood or any face garments.
directly facing the camera and only one !erson can be !rocessed in a
given time
%(cluding !hotos taken outdoors and ith bad eather conditions
such as heavy rain or foggy environment. /oo much sunlight e(!osure
-
7/24/2019 sample thesis defense ppt
4/26
Methodology
0ata $athering
1hat2s our data to su!!ort our study3 &ili!ino face images of +- individuals
1hat tool did e used to get these face images3 built'in la!to! ebcam of )enovo y45 ith a resolution of *+(+- at .6 mega!i(els
-
7/24/2019 sample thesis defense ppt
5/26
Methodology
0ata $athering &lo : "etu!
)a!to! ith built'in ebcam is !laced on a table
&or sitting !osition a chair is !rovided
1hite background
7eutral'lighted room
-
7/24/2019 sample thesis defense ppt
6/26
Methodology
0ata $athering &lo : 8rocedure
"ubject as instructed to sit don on the
chair
"ubject as guided to !osition face and look
in front of the eb camera 6+ inches aay
from it
"ubject as requested to follo the rules on
taking !hoto of him9her
-
7/24/2019 sample thesis defense ppt
7/26
Methodology
$eneral 8rocess "te! 6: %(!erimental Algorithm /esting;
7oise #eduction
est est "u!er #esulotion ?@'+4@'64@;
-
7/24/2019 sample thesis defense ppt
8/26
Adding Gaussian Noise
-
7/24/2019 sample thesis defense ppt
9/26
Noise Reduction Testing (Fuzzy Filter,Guassian with Edge Preservation,Median Filter)
-
7/24/2019 sample thesis defense ppt
10/26
ME and PNR testing
-
7/24/2019 sample thesis defense ppt
11/26
Methodology
"te! 6: 8re!rocessing 7ormali=ation Cro!!ing and resi=ing;
&our face !oints that ere used in cro!!ing the image:
/o! of the hair, left ear, right ear, ti! of the chin
After cro!!ing, resolution of the images as converted to 5(6+5 !i(els
-
7/24/2019 sample thesis defense ppt
12/26
Methodology
"te! 6: 8re!rocessing 7oise #eduction &u==y &ilter;
#$> values of the image ill undergo &u==y logic rules
-
7/24/2019 sample thesis defense ppt
13/26
Methodology
"te! 6: 8re!rocessing 7oise #eduction; %ach !i(el is com!osed of #$> values +-,B,6B;
Assign a membershi! value for each color !i(el &or vertical a(is:
,B4D E : 7on noisy
B?,644D E .6'6: 8robably noisy
64?,+44D E 6: 7oisy !i(el
&or Fori=ontal a(is: 0ivide the !i(el value by +4?
Classify hether dark medium or light
&u==y #ules
-
7/24/2019 sample thesis defense ppt
14/26
Methodology
"te! 6: 8re!rocessing rightness adjustment increase the total contrast of the image
-
7/24/2019 sample thesis defense ppt
15/26
Methodology
"te! +: Main 8rocess 8hase 6 E 8i(el #atio %valuation
8hase + E Actual #esults and /esting
Image Database
Phase 1
Parameter Testing Set
+ lo resolution in!uts ill be used for the 4@ !i(el
ratio
+ lo resolution images ill be included in the face
database for the 6@ and -@ !i(el ratio
Phase 2
Main Process Set
+ lo resolution in!uts ill be used for the 4@ !i(el
ratio 0ifferent from the !revious set;
+ lo resolution images ill be included in the face
database for the 6@ and -@ !i(el ratio
-
7/24/2019 sample thesis defense ppt
16/26
Methodology
"te! +: Main 8rocess 8hase 6 E 8i(el #atio %valuation
!"#T$!%(M)& !(M) & !'(M)
"
-
7/24/2019 sample thesis defense ppt
17/26
Methodology
"te! +: Main 8rocess 8hase 6 E 8i(el #atio %valuation
-
7/24/2019 sample thesis defense ppt
18/26
Methodology
"te! +: Main 8rocess 8hase 6 E 8i(el #atio %valuation
GOH/6IG
-
7/24/2019 sample thesis defense ppt
19/26
Methodology
"te! +: Main 8rocess 8hase 6 E 8i(el #atio %valuation
M"% and 8"7# Calculation
&acial &eature %(traction
-
7/24/2019 sample thesis defense ppt
20/26
Methodology"te! +: Main 8rocess
8hase 6 E #esults
Pixel Ratio (M) Unprocesse
d Image
Processed
Image
MSE
PSNR
Feature
Extraction
Input = 5!
Similar Images =
"!#ig$est Similar
= %!
*+
.*...
Input = 5!
Similar Images =
&!
#ig$est Similar
= '!
./0*
.1*//1
Input = 5!
Similar Images =
'!
#ig$est Similar
= &!
./23*+
./*+.0
-
7/24/2019 sample thesis defense ppt
21/26
Methodology"te! +: Main 8rocess
8hase 6 E "ummari=ed #esults
8arameter 4'6'- yielded highest overall detection rate getting 54@
mouth detection
)oest M"%
Fighest 8"7#Pixel Ratio (M)
E*e
R
E*e
Nose Mout$ +,e
MSE
+,e
PSNRInput = 5!
Similar Images =
"!
#ig$est Similar =
%!
.++
4
.++4 +4 314 .22.* .*31
Input = 5!
Similar Images =
&!
#ig$est Similar =
'!
.++
4
.++4 314 14 .0*+1 .1*1.
Input = 5!
Similar Images =
'!
#ig$est Similar =
&!
.++
4
.++4 +4 14 .+*02 .*.
-
7/24/2019 sample thesis defense ppt
22/26
Methodology"te! +: Main 8rocess
8hase + E Actual #esults and /esting GOH/6IG).4; Gs.6; GF.-;
8erformed on another + images
-
7/24/2019 sample thesis defense ppt
23/26
Methodology"te! +: Main 8rocess
8hase + E Actual #esults and /esting &acial &eature %(traction Algorithm
-
7/24/2019 sample thesis defense ppt
24/26
Methodology"te! +: Main 8rocess
8hase + E Actual #esults and /esting #esults "ummary
.$ole Face e/t E*e Rig$t E*e Nose Mout$
Unprocesse
d
14
(. out o5 2+
i6ages)
+4
(.3 out o5 2+
i6ages)
14
(.1 out o5 2+
i6ages)
14
(. out o5
2+ i6ages)
.14
( out o5 2+
i6ages)
Processed .++4
(2+ out o5 2+
i6ages)
.++4
(2+ out o5 2+
i6ages)
.++4
(2+ out o5 2+
i6ages)
.++4
(2+ out o5
2+ i6ages)
2+4
(0 out o5 2+
i6ages)
Impro,eme
nt
1*2/4 ..*..4 *4 1*2/4 *4
-
7/24/2019 sample thesis defense ppt
25/26
Conclusion
#esults shoed that the combination of the algorithms significantly
im!roved the quality of face images taken from a lo resolution
camera by getting !ositive im!rovement on facial feature detection
4.+?@ facial e(traction,
66.66@ left eye **.**@ right eye
4.+?@ nose
**.**@ mouth
-
7/24/2019 sample thesis defense ppt
26/26
#ecommendation A larger face database may !rove to be useful for better image
reconstruction etter data gathering setu! for face database considering subjects
orientation and !ositioning