SlideShare a Scribd company logo
Image Processing Report
Histogram Equalization
Image Processing - Report #1
Due date: April. 2, 2018
In our class, we have learned about a histogram equalization
technique to enhance the images's quality. Histograms are the basics
for numerous spatial domain techniques. Simple examples can help
you to understand how to equalize the histogram in an image.
Suppose that Fig. 1 is a 4-bit input image(L=16) of size 6×6 pixels.
And also we assume that 0 is black and 15 is white.
Now calculate the histogram equalized image. You need to calculate
histogram, PDF(probability density function), CDF(cumulative density
function), and the table for converting  to . When you calculate s
values, , you round them to the nearest integer. Finally, explain the
result comparing to the input image in point of their contrasts. You
have to submit the report in class.
1. The Histogram of the input image.
 0 1 2 3 4 5 6 7
frequency 13 12 6 3 2 0 0 0
 8 9 10 11 12 13 14 15
frequency 0 0 0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
1 1 1 1 1 1
2 2 2 2 2 2
1 1 1 1 1 1
3 3 3 4 4 0
Fig.1 Input image
2. PDF(probability density function)
3. CDF(cumulative density function)
4. the table for converting  to  ( (prev Expression) * (L-1) )
5. The histogram equalized image
 0 1 2 3 4 5 6 7
   13/36 12/36 6/36 3/36 2/36 0 0 0
 8 9 10 11 12 13 14 15
   0 0 0 0 0 0 0 0
 0 1 2 3 4 5 6 7
   13/36 25/36 31/36 34/36 1 1 1 1
 8 9 10 11 12 13 14 15
   1 1 1 1 1 1 1 1
 0 1 2 3 4 5 6 7

13/36
*15
25/36
*15
31/36
*15
34/36
*15
15 15 15 15
 8 9 10 11 12 13 14 15
 15 15 15 15 15 15 15 15
5.4 5.4 5.4 5.4 5.4 5.4
0 0 0 0 0 0
0 0 0 0 0 0
1 1 1 1 1 1
2 2 2 2 2 2
1 1 1 1 1 1
3 3 3 4 4 0
Fig.1 Input image
가장 근접한 정수로 반올림
6. Explain the result comparing to the input image in point of their
contrasts.
을 통해 입력 의 저대비 및 낮은 평균Histogram Equalization Matrix Gray
의 값의 영상을 비교적 고른 분포의 을 가진 로 바꿀Level Histogram Matrix
수 있었다.
하지만 변환을 수행한 표만 가지고 비교함은 가식성이 떨어지기 때문에 위,
의 풀이과정을 알고리즘화 하여 코드로 변환한 후 주어진 를 가지고, Matrix
5.4 5.4 5.4 5.4 5.4 5.4
10.4 10.4 10.4 10.4 10.4 10.4
12.9 12.9 12.9 12.9 12.9 12.9
10.4 10.4 10.4 10.4 10.4 10.4
14.1 14.1 14.1 15 15 5.4
Fig.2 Output image
5 5 5 5 5 5
5 5 5 5 5 5
10 10 10 10 10 10
13 13 13 13 13 13
10 10 10 10 10 10
14 14 14 15 15 5
Fig.3 Output image with rounded
 0 1 2 3 4 5 6 7
frequency 0 0 0 0 0 13 0 0
 8 9 10 11 12 13 14 15
frequency 0 0 12 0 0 6 3 2
Fig.4 Histogram of the Output image
연산을 진행하였다Histogram Equalization .
전 후 의 의 빈도수를 비교했을 때 기존의 대비, Matrix Gray Level Matrix
균일해졌다고 볼 수 있다.
하지만 주어진 의 이 으로 작기 때문에Matrix shape (6, 6) Histogram
을 통해 개선된 여부를 쉽게 파악하기가 쉽지 않다Equalization (Enhanced) .
따라서 조금 더 적절한 예시로 파일을 활용하여 같은 알고리즘을, lena.png
이용해 을 적용한다Histogram Equalization .
결과는 다음과 같다.
기존 그리고 영역에 비어있는 의 빈도수를0~50, 225~250 Gray Level
을 통해 전체 영역에 빈도수가 고르게 분포Equalization 0~255 Gray Level
된 것 고대비 영상으로 변한 것 을 확인할 수 있었다( ) .
또한 을 통해 도출된 값들을 합하여 의, Histogram Equalization Matrix Size
로 나눈값이
의 절반값과 비슷해 진다는 것을 알 수 있었다Max Gray Level(2^Bit 1) .–
입력 행렬에서1) 6*6 ,
(5*13 + 10*12 + 13*6 + 14*3 +15*2) / (6*6) = 9.3
16/2 = 8
에서2) lena.png ,
33743418/ (512*512) = 128.7209
256 / 2 = 128
좀 더 큰 배열을 입력으로 주었을 때 그 근사정도가 커졌다고 볼 수 있다.
따라서 다음과 같은 식을 도출할 수 있었으며 에 대한 근사정도가 커질, , L/2
수록 의 빈도수가 고르게 분배되었 다는 것을 알 수 있었다Gray Level .
  
 
  __≈

More Related Content

PDF
[기초개념] Graph Convolutional Network (GCN)
PDF
[Paper] eXplainable ai(xai) in computer vision
PDF
Lecture 3: Unsupervised Learning
PPTX
Dsh data sensitive hashing for high dimensional k-nn search
PDF
Lecture 4: Neural Networks I
PPTX
Generative adversarial network
PPTX
Variational AutoEncoder(VAE)
PPTX
111118 ch 4_basic image manipulation_web
[기초개념] Graph Convolutional Network (GCN)
[Paper] eXplainable ai(xai) in computer vision
Lecture 3: Unsupervised Learning
Dsh data sensitive hashing for high dimensional k-nn search
Lecture 4: Neural Networks I
Generative adversarial network
Variational AutoEncoder(VAE)
111118 ch 4_basic image manipulation_web

What's hot (20)

PPTX
Recurrent Neural Net의 이론과 설명
PDF
Lecture 2: Supervised Learning
PDF
Variational Auto Encoder, Generative Adversarial Model
PPTX
GAN with Mathematics
PPTX
MNIST for ML beginners
PPTX
Survey on Monocular Depth Estimation
PDF
차원축소 훑어보기 (PCA, SVD, NMF)
PDF
06월 26일 진로 3 d 프로그래밍 발표
PPTX
03.12 cnn backpropagation
PPTX
Les net
PPTX
Visualizing data using t-SNE
PDF
Optimization algorithms in machine learning
PPTX
Graph Neural Network (한국어)
PPTX
neural network 기초
PDF
CycleGAN이 무엇인지 알아보자
PDF
Vs^3 net for machine reading comprehension question answering
PPTX
Data Visualization and t-SNE
PDF
오픈드론맵 한국어 사용자 지침서
PDF
Big data 시대의 통계학
PPTX
Deep learning seminar_snu_161031
Recurrent Neural Net의 이론과 설명
Lecture 2: Supervised Learning
Variational Auto Encoder, Generative Adversarial Model
GAN with Mathematics
MNIST for ML beginners
Survey on Monocular Depth Estimation
차원축소 훑어보기 (PCA, SVD, NMF)
06월 26일 진로 3 d 프로그래밍 발표
03.12 cnn backpropagation
Les net
Visualizing data using t-SNE
Optimization algorithms in machine learning
Graph Neural Network (한국어)
neural network 기초
CycleGAN이 무엇인지 알아보자
Vs^3 net for machine reading comprehension question answering
Data Visualization and t-SNE
오픈드론맵 한국어 사용자 지침서
Big data 시대의 통계학
Deep learning seminar_snu_161031
Ad

More from 우진 신 (15)

PDF
비교논문 분석 - 디지털 IoT 도어락 DDiT
PPTX
INSTAGRatio 소개영상 (instagram 서드파티 어플리케이션)
PDF
Windows API Programming Project Report
PDF
Mlp logical input pattern classfication report doc
PDF
171220 웹프로그래밍 web app 토렌트 관리체계
PPTX
주차관리 시스템 프로토타입 제안서
PPTX
유비쿼터스 IoT 도어락 프로토타입 발표자료
PPTX
유비쿼터스 Gps원리 발표자료
PPTX
데이터베이스(상) 수업 요구분석 및 모델링 발표자료
PPTX
데이터베이스(상) 수업 요구분석 및 Er d 변형 2차 발표자료
PPTX
데이터베이스(상) 수업 요구분석 및 Er d 변형 1차 발표자료
PPTX
데이터베이스(상) 12년도 da설계 공모대전 풀이
PPTX
사물인터넷을 활용한 디지털 도어락 DDiT 비교논문 발표자료 1차
PPTX
171105 웹프로그래밍 static web 몰빵 발표자료
PPTX
C# 재무관리 프로그램(주제고정) 발표자료
비교논문 분석 - 디지털 IoT 도어락 DDiT
INSTAGRatio 소개영상 (instagram 서드파티 어플리케이션)
Windows API Programming Project Report
Mlp logical input pattern classfication report doc
171220 웹프로그래밍 web app 토렌트 관리체계
주차관리 시스템 프로토타입 제안서
유비쿼터스 IoT 도어락 프로토타입 발표자료
유비쿼터스 Gps원리 발표자료
데이터베이스(상) 수업 요구분석 및 모델링 발표자료
데이터베이스(상) 수업 요구분석 및 Er d 변형 2차 발표자료
데이터베이스(상) 수업 요구분석 및 Er d 변형 1차 발표자료
데이터베이스(상) 12년도 da설계 공모대전 풀이
사물인터넷을 활용한 디지털 도어락 DDiT 비교논문 발표자료 1차
171105 웹프로그래밍 static web 몰빵 발표자료
C# 재무관리 프로그램(주제고정) 발표자료
Ad

Image processing - Histogram Equalization

  • 2. Image Processing - Report #1 Due date: April. 2, 2018 In our class, we have learned about a histogram equalization technique to enhance the images's quality. Histograms are the basics for numerous spatial domain techniques. Simple examples can help you to understand how to equalize the histogram in an image. Suppose that Fig. 1 is a 4-bit input image(L=16) of size 6×6 pixels. And also we assume that 0 is black and 15 is white. Now calculate the histogram equalized image. You need to calculate histogram, PDF(probability density function), CDF(cumulative density function), and the table for converting  to . When you calculate s values, , you round them to the nearest integer. Finally, explain the result comparing to the input image in point of their contrasts. You have to submit the report in class. 1. The Histogram of the input image.  0 1 2 3 4 5 6 7 frequency 13 12 6 3 2 0 0 0  8 9 10 11 12 13 14 15 frequency 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 2 2 2 2 2 2 1 1 1 1 1 1 3 3 3 4 4 0 Fig.1 Input image
  • 3. 2. PDF(probability density function) 3. CDF(cumulative density function) 4. the table for converting  to  ( (prev Expression) * (L-1) ) 5. The histogram equalized image  0 1 2 3 4 5 6 7    13/36 12/36 6/36 3/36 2/36 0 0 0  8 9 10 11 12 13 14 15    0 0 0 0 0 0 0 0  0 1 2 3 4 5 6 7    13/36 25/36 31/36 34/36 1 1 1 1  8 9 10 11 12 13 14 15    1 1 1 1 1 1 1 1  0 1 2 3 4 5 6 7  13/36 *15 25/36 *15 31/36 *15 34/36 *15 15 15 15 15  8 9 10 11 12 13 14 15  15 15 15 15 15 15 15 15 5.4 5.4 5.4 5.4 5.4 5.4 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 2 2 2 2 2 2 1 1 1 1 1 1 3 3 3 4 4 0 Fig.1 Input image
  • 4. 가장 근접한 정수로 반올림 6. Explain the result comparing to the input image in point of their contrasts. 을 통해 입력 의 저대비 및 낮은 평균Histogram Equalization Matrix Gray 의 값의 영상을 비교적 고른 분포의 을 가진 로 바꿀Level Histogram Matrix 수 있었다. 하지만 변환을 수행한 표만 가지고 비교함은 가식성이 떨어지기 때문에 위, 의 풀이과정을 알고리즘화 하여 코드로 변환한 후 주어진 를 가지고, Matrix 5.4 5.4 5.4 5.4 5.4 5.4 10.4 10.4 10.4 10.4 10.4 10.4 12.9 12.9 12.9 12.9 12.9 12.9 10.4 10.4 10.4 10.4 10.4 10.4 14.1 14.1 14.1 15 15 5.4 Fig.2 Output image 5 5 5 5 5 5 5 5 5 5 5 5 10 10 10 10 10 10 13 13 13 13 13 13 10 10 10 10 10 10 14 14 14 15 15 5 Fig.3 Output image with rounded  0 1 2 3 4 5 6 7 frequency 0 0 0 0 0 13 0 0  8 9 10 11 12 13 14 15 frequency 0 0 12 0 0 6 3 2 Fig.4 Histogram of the Output image
  • 5. 연산을 진행하였다Histogram Equalization . 전 후 의 의 빈도수를 비교했을 때 기존의 대비, Matrix Gray Level Matrix 균일해졌다고 볼 수 있다. 하지만 주어진 의 이 으로 작기 때문에Matrix shape (6, 6) Histogram 을 통해 개선된 여부를 쉽게 파악하기가 쉽지 않다Equalization (Enhanced) . 따라서 조금 더 적절한 예시로 파일을 활용하여 같은 알고리즘을, lena.png 이용해 을 적용한다Histogram Equalization . 결과는 다음과 같다.
  • 6. 기존 그리고 영역에 비어있는 의 빈도수를0~50, 225~250 Gray Level 을 통해 전체 영역에 빈도수가 고르게 분포Equalization 0~255 Gray Level 된 것 고대비 영상으로 변한 것 을 확인할 수 있었다( ) . 또한 을 통해 도출된 값들을 합하여 의, Histogram Equalization Matrix Size 로 나눈값이 의 절반값과 비슷해 진다는 것을 알 수 있었다Max Gray Level(2^Bit 1) .– 입력 행렬에서1) 6*6 , (5*13 + 10*12 + 13*6 + 14*3 +15*2) / (6*6) = 9.3 16/2 = 8 에서2) lena.png , 33743418/ (512*512) = 128.7209 256 / 2 = 128 좀 더 큰 배열을 입력으로 주었을 때 그 근사정도가 커졌다고 볼 수 있다.
  • 7. 따라서 다음과 같은 식을 도출할 수 있었으며 에 대한 근사정도가 커질, , L/2 수록 의 빈도수가 고르게 분배되었 다는 것을 알 수 있었다Gray Level .        __≈