A Robust Single Image Dehazing Method
The visibility of outdoor scenes is degraded by weather phenomena such as mist, fog and rain. The degradation is due to the substantial presence of atmospheric particles that scatter and absorb light. Poor visibility in bad weather conditions is considered a major problem for many applications of co...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | en_US |
| Subjects: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The visibility of outdoor scenes is degraded by weather phenomena such as mist, fog and rain. The degradation is due to the substantial presence of atmospheric particles that scatter and absorb light. Poor visibility in bad weather conditions is considered a major problem for many applications of computer vision such as surveillance, intelligent vehicles, and outdoor object recognition. To solve the problem of degraded visibility, many enhancement and restoration methods have been proposed. Some of these methods require additional information such as using two images that are captured in different weather conditions, whereas other methods focus on recovering scene albedo of a single image by using different assumptions. However, the main drawback of enhancement techniques is enhancing the quality of the degraded images without considering the degradation level of visibility. In addition, the main limitation of the methods that use global atmospheric light (GAL) to restore the degraded images is the inaccurate estimation of GAL. This thesis proposed a scheme that consists of three phases to solve the problem of outdoor degraded images. In phase one, a method called ICLAHE is suggested to enhance the intensity of low visibility by introducing new clip limit parameter of Contrast Limited Adaptive Histogram Equalization technique that can vary based on the type of weather conditions. The clip limit mainly controls the degree of the contrast of the enhanced image. Phase two proposed an algorithm to tackle the problem of high intensity pixels that belong to white objects that lead to inaccurate estimation of GAL. The proposed algorithm is called EGAL. In the third phase, a method called RIGAL is proposed to remove the effects of atmospheric particles from outdoor degraded images by exploiting the results of the previous phases. It recovers the scene albedo while maintaining the fine details. Testing results have shown that phase one enhances the degraded intensity to be close to 4% to the intensity of the free haze scene under different level of visibility and haze density. In addition, the findings indicate that phase two estimated GAL accurately to reach 0.1% close to the manual reference of GAL estimation under different weather conditions. Furthermore, results of phase three proved that the sufficiency of removing the effects of atmospheric particles with preserving fine details reaches to 2% of the quality of the free haze image against state of the art of enhancement and restoration techniques. |
|---|