Enhanced dark channel prior and transmission map estimation techniques for removing foreground dense haze on static image
Haze removal on a degraded image was the challenging task in image processing field. A reliable technique must be able to remove dense haze effects on the static image, in addition, improving the quality of the image. Hence, this research proposes the integration between the two enhanced techniques...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English English |
| Published: |
2019
|
| Subjects: | |
| Online Access: | https://eprints.ums.edu.my/id/eprint/41358/2/24%20PAGES.pdf https://eprints.ums.edu.my/id/eprint/41358/1/FULLTEXT.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Haze removal on a degraded image was the challenging task in image processing field. A reliable technique must be able to remove dense haze effects on the static image, in addition, improving the quality of the image. Hence, this research proposes the integration between the two enhanced techniques for dense haze effects restoration at the foreground of outdoor scenery image. For the first technique of modified dark channel prior allows it to remove haze, which improves of contrast quality and obtains the haze intensity. The intensity of dark channel allows estimation of the thickness value of haze. Additionally the second technique, transmission map estimation, was modified to improve the image colour quality and its details. Both of these techniques will be integrated to form a full process of dehazing technique. The restored image then proceeds to the post-processing stage for contrast enhancement and gamma correction. Histogram equalization technique is used for contrast enhancement and gamma correction is used for colour correction of image, with improved brightness. Besides that, the median filter is applied for transmission refinement and to improve edges on the image scene. Several tests were implemented in order to validate the results of this experiment. For example, the Mean Square Error (MSE) test and Peak Signal to Noise Ratio (PSNR) test were used for the image quality assessment. The obtained test values show that the proposed technique achieved better results with lower values for MSE and higher values for PSNR compared to the other established haze removal techniques in the field of image enhancement; Kaiming and Gibson techniques. The MSE test value for the proposed technique was only 2009.76dB, meanwhile, Kaiming and Gibson techniques showed 3920.13dB and 2785.07dB respectively and the PSNR test value was 15.13dB, however, both of Kaiming and Gibson techniques recorded 12.23dB and 13.72dB respectively which further proves that the proposed technique produces a better image. Finally, a test of Structural Similarity Index (SSIM) will be taken in order to show that the output scene of haze-free image was not similar to the input scene of hazy image. In general, the proposed technique is able to remove the foreground dense haze from the outdoor scenery static images and at the same time improves the contrast and colour quality on-scene image. |
|---|