Hybrid denoising framework for pneumonia detection: integrating discrete wavelet transform and Pix2Pix GAN

Pneumonia is a respiratory illness that manifests in fluid or pus accumulation by inflaming the lungs’ air sacs. Coughing, fever, and difficulty breathing are some of the signs of this illness, which interrupts the regular flow of oxygen. Serious problems may only be avoided with a correct diagnosis and effective treatment. This study improves pneumonia diagnosis using Chest X-ray Images (CXIm) by employing an innovative image-denoising technique. This innovation reduces noise without compromising structural integrity, thereby exposing significant features. The hybrid framework denoises CXIm with DWT/Pix2Pix GAN. DWT breaks the noisy CXIm image into discrete frequency components. This reduces noise and preserves image structure. Pix2PixGAN deconstructs and clarifies denoised images. Experiments with DWT-Pix2PixGAN increase CXIm quality. A higher SSIM of 0.8721 indicates better structure retention, while 33.0230 dB and 18.5227 dB indicate noise reduction. Lower MSE of 0.9412 indicates better reconstruction accuracy. This study found that DWT/Pix2PixGAN, a hybrid denoising, improves CXIm. This method improves image clarity and structure retention over conventional denoising. This method works because SSIM, SNR, PSNR, and MSE improved. It minimises noise and improves diagnostic accuracy, making it suitable for pneumonia diagnosis. Research may improve and expand the use of additional medical imaging modalities.