Image Enhancement Using Deblur Generative Network and Deep Deblur Adversarial

The increased use of digital cameras has given rise to new challenges such as reduced image quality resulting in poor clarity or distortion, which can result in loss or unclearness of recorded information. To overcome this problem, deep learning-based approaches such as DeblurGAN and DeepDeblur have been developed for the clarity of blurry images. This research aims to compare the performance of the two approaches in the context of improving the clarity of blurry images. The research steps include collecting data from the GoPro dataset, dividing the data into training data and test data, analyzing the data using the DeblurGAN and DeepDeblur methods, and evaluating the results using Structural Similarity Index Measure (SSIM), Mean Square Estimation (MSE), and Peak Signal-to-Noise Ratio (PSNR). The evaluation results show that the DeblurGAN method achieves the best results with SSIM values of 0.958, MSE of 0.00057, and PSNR of 33.18. This shows that DeblurGAN has better evaluation performance compared to Deep Deblur in improving the clarity of blurry images.