Image Denoising Using Wavelet Transform and Genetic Programming

This research develops a superior image denoising approach through combining Discrete Wavelet Transform (DWT) with Genetic Programming (GP) for achieving outstanding noise reduction performance. The initial stage of the method includes decomposing noisy input images through DWT into two distinct parts: low-frequency approximation and high-frequency detail sub-bands. The majority of noise appears in the high-frequency divisions within the spectrum. The inability of traditional thresholding methods to adapt to noise variations due to different environments explains why Genetic Programming contributes customized thresholding expressions. The thresholds created through GP adaptation method process wavelet coefficients in a way that eliminates noise without damaging essential structural information. The process of inverse wavelet transform generates visually appealing images with reduced noise levels. Standard image dataset experiments confirm the proposed approach delivers better results than general denoising techniques. The quantitative evaluation based on PSNR, SSIM and MSE metrics shows significant enhancement of numerical and visual performance metrics. The hybrid framework demonstrates excellent capabilities to reduce noise while maintaining image details which positions it as an excellent solution for image restoration needs in medical diagnostics and remote sensing and digital media processing.