ULTRASOUND AND THERMAL IMAGE ENHANCEMENT TECHNIQUE USING CONVOLUTION NEURAL NETWORK

Medical imaging is a non-invasive diagnosis technique in the field of medical imaging with non-radioactive nature. However, the efficiency of medical image (thermal and ultrasound) reconstruction affected due to noisy nature, bulk size and time consumption. Therefore, to get high quality reconstructed image we have presented sparsity based image enhancement (SIE) technique using convolution neural network (CNN) namely SIE-CNN for eliminating noise and improving reconstruction quality of thermal and ultrasound image. The proposed image enhancement technique adopted fast CCN architecture (the CNN architecture aid in training larger dataset with improved quality) and sparsity based image enhancement technique is used for building template learning (DL) technique for removing noise. Experiment are conducted to evaluate performance of SIE-CNN over existing model in terms of PSNR and SSIM. The result attained shows the proposed SIE-CNN model achieves better PSNR and SSIM performance than existing image enhancement technique. Thus, SIE-CNN achieves much better reconstruction quality.