Histogram-Based Transformation Function Estimation for Low-Light Image Enhancement

Park, Jaemin; Vien, An Gia; Kim, Jin Hwan; Lee, Chul

doi:10.1109/icip46576.2022.9897778

Cited by 7 publications

(4 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Park et al, proposed a learning-based low-light image enhancement algorithm, a histogram-based transform function estimation network that estimates transform functions using the histogram of an input image. This method was applied to low-light image enhancement using channel-wise intensity transform to obtain enhanced images (Park et al, 2022). Liu et al, proposed an enhancement method for low-light Unmanned Aerial Vehicle images.…”

Section: Literature Reviewmentioning

confidence: 99%

Aircraft Accident and Crash Images Processing with Machine Learning

Gümüş,

Dursun

2024

Journal of Aviation

View full text Add to dashboard Cite

The aviation industry is in constant need of innovations in terms of safety and operational efficiency. In this context, low-light image enhancement technologies play an important role in a numerous areas of disciplines, from night flights to accident and collision investigations. Machine learning, deep learning methods and traditional methods not only provide the aviation industry with an effective image processing and improvement capacity in low light conditions, but also reveal important information by analysing the data of low-light images of crashed and destroyed aircraft. Within the scope of the study, traditional methods, deep learning method and machine learning are combined in order to enhance and process low-light ambient images of crashed and destroyed aircraft. By using Swish and Tanh activation functions together in the deep learning model, the performance of the neural networks used in the process of improving low-light environment images was improved and the image quality was increased. The enhanced images were evaluated and compared using PSNR and MSE as objective quality assessment measures. According to the PSNR and MSE criteria, the numerical results obtained from the image enhancement studies of the deep learning model were calculated as 29.85 and 100.44, respectively. The results introduce that the deep learning model provides better image enhancement than traditional methods. In conclusion, improvement of low-light image and processing is an important technological advancement in the aviation industry, enabling safer and more efficient operations. The successful of machine learning include deep learning and traditional methods shows that the aviation industry will achieve a safer and innovative structure in the future.

show abstract

Section: Literature Reviewmentioning

confidence: 99%

Aircraft Accident and Crash Images Processing with Machine Learning

Gümüş,

Dursun

2024

Journal of Aviation

View full text Add to dashboard Cite

show abstract

“…However, acquiring ISAR data is typically more expensive and complex compared to obtaining optical image data or Proposed (40.14, 96) RIFE [2] (32.56, 93) DMVFN [3] (36.83, 96) MTFE [4] (34.27, 81) UPR-Net [5] (37.83, 96)…”

Section: Introductionmentioning

confidence: 99%

“…Our proposed method achieves best for both of the two metrics compared with other methods, including optical flow-based[2]-[5] and GAN-based[6] approaches. Specifically, compared with the traditional image-to-image translation method Pix2Pix GAN[6] and the latest flow-based interpolation method MTFE[4], PSNR and SSIM of the results interpolated by our proposed method has increased by 12.49 dB, 5.87 dB, and 22%, 15% respectively. All comparative approaches are conducted on a single RTX 3080 GPU.…”

mentioning

confidence: 99%