Author(s): Ahmed Fraidoon Abdulkarem

Abstract: The Deep Detail Network (DDN) was created to eliminate rain streaks from photographs. Deep convolutional neural networks (CNN) are taught with data to learn how to correlate rainy and clean feature layers in pictures. We create our own rainy photos for training because we do not have access to real-world ones. BCET enhances contrast in low-light pictures by modifying the objective function and adjusting contrast, rather than simply increasing the network's depth or breadth. We trained our Deep Neural Network (DDN) using two datasets of rainy photos. Testing DDN on actual images revealed that the network achieved good performance, even though it was trained on artificial data. Image processing is utilized to enhance the visual outputs of the CNN architecture. Improvements in rain removal include enhancing the contrast in wet and low-light photographs and increasing the computational speed after implementing a new network, surpassing previous approaches.

DOI: 10.33545/27076571.2024.v5.i1a.80

Pages: 30-37 | Views: 116 | Downloads: 55

Download Full Article: Click Here