Enhanced Single Remote Sensing Image Dehazing via Vision Transformer with Silencing Map Transmission and advanced Image Processing Techniques
Sivasubramanyam Medasani, B. Shireesha, B. Mallikarjun Reddy, CH. Charan Sai Varaha Teja, E. Venkata Sainath Chowdary · 2024
- This research provides an improved single remote sensing picture dehazing method using sophisticated image processing techniques and Vision Transformer with Silencing Map Transmission. The dehazing output generation, morphological analysis, dark channel prior, silencing map transmission, recurrent neural networks (RNN), and input image processing are some of the components that are included in the suggested method. The objective is to far outperform current techniques in terms of dehazed image quality. The initial action in the process is getting ready the input image, after which morphological analysis is performed to determine the properties of the haze. The key characteristics of the haze are retrieved by using the dark channel before. Transmitting silencing maps improves our comprehension of scene depth, which is important for precise dehazing. RNN is used to learn from sequential data and propagate information effectively. Clearer and more aesthetically pleasing photos are produced by the dehazing output, which is generated based on the learned attributes. Mean Squared Error (MSE), Peak Signal-to-Noise Ratio (PSNR), and Structural Similarity Index Measure (SSIM), and Time of Execution (TOE) are among the metrics used to assess the efficacy of the suggested approach. The supremacy of the suggested strategy is shown in contrast to current approaches. The trial results demonstrate considerable improvements in PSNR, SSIM, and TOE measures across a variety of test images. In particular, the suggested approach outperforms current methods in terms of PSNR improvements of up to 4.23 dB and TOE reductions of up to 2.58 seconds. To sum up, the enhanced single remote sensing picture dehazing approach that has been described provides significant improvements in both image quality and processing efficiency. As such, it is a promising solution that might be used for a variety of remote sensing applications.