| URN |
etd-0016117-121822 |
|---|
| Author |
Chia-ting Kuo |
|---|
| Author's Email Address |
No Public. |
|---|
| Statistics |
This thesis had been viewed 5575 times. Download 18 times. |
|---|
| Department |
Computer Science and Engineering |
|---|
| Year |
2016 |
|---|
| Semester |
1 |
|---|
| Degree |
Master |
|---|
| Type of Document |
|
|---|
| Language |
zh-TW.Big5 Chinese |
|---|
| Title |
Underwater Image Dehazing and Enhancement by Compensating Light Absorption Loss and Multi-exposure Fusion |
|---|
| Date of Defense |
2017-01-06 |
|---|
| Page Count |
64 |
|---|
| Keyword |
Image Dehazing
Underwater Image
Multi-exposure Fusion
Image Enhancement
Color Cast
Dark Channel Prior
|
|---|
| Abstract |
In recent years, the topic of single image dehazing has received a lot of attention, and many enhancing algorithms have been proposed to efficiently alleviate the effect of light scattering. However, these algorithms cannot be directly applied to the underwater images effectively. One of the main reasons is that the color cast caused by various degrees of absorption for different light wavelengths in the underwater environment cannot be neglected. Therefore, this thesis proposes a series of image processing steps in order to enhance the color contrast and correct the color cast of the underwater images. Our first step is to apply Red Channel Prior approach with modified pixel selection criteria in order to estimate the underwater background light intensity. In order to obtain the transmission map without halo artifacts, weighted least square filter has been used to refine the map to preserve the edge. After eliminating the scattering attenuation effect, our next step is to compensate the absorption loss along the propagation path between objects and camera. Then, the white balance algorithm is utilized to remove the color cast caused by the light attenuation along the propagation path between the water surface and the objects. Finally, multiple images obtained by histogram stretching of different scales will be fused to adjust the light intensity of the dehazed images. Our experimental results show that the proposed approach can achieve the best quantitative visual metric results for most of the test images compared with those recent state-of-the-art ones. We also conduct SIFT feature test in order to illustrate if our enhanced method can help increasing the image features. Our test results show that an average of 40 features can be found matched after applying our dehazing method compared with only four can be detected in the original images. |
|---|
| Advisory Committee |
Tang-Kai Yin - chair
Shiann-Rong Kuang - co-chair
Yun-Nan Chang - advisor
|
|---|
| Files |
Indicate in-campus at 3 year and off-campus access at 3 year. |
|---|
| Date of Submission |
2017-01-16 |
|---|
