A Hybrid Wavelet-Based Restoration Method for Agricultural Images

Abstract

To improve denoising results for agricultural images, a hybrid wavelet-based method of agricultural image denoising was developed that was inspired by the idea of crossbreeding in living beings and the crossover operator in genetic algorithms. This method hybridized the advantages of two individual image-denoising methods using the crossover and mutation operators. The noisy image was denoised via wavelet thresholding and by applying an additional filter (such as a Wiener filter or a median filter), and the two denoised images were called the male and female parents. Next, the two parents were transformed into the encoding space to form the chromosomes. Then, the chromosomes were selected via the fitness function using the selection and crossover rates for the crossover and mutation operators. After several generations of crossover and mutation operations, the superior offspring was obtained. Finally, the offspring was decoded as an image with better noise removal results. The effectiveness of this proposed algorithm was tested using both agricultural images (wheat and apple) and a non-agricultural image (Lena). The experimental results indicated that this new denoising algorithm achieved better performance than conventional denoising methods in terms of visual quality and peak signal-to-noise ratio (PSNR). The original PSNR values of the three images were 59.22, 59.63, and 59.67 dB, whereas the PSNR values were 73.26 and 71.62 dB for the wheat and apple images using the hybrid wavelet-average method, and the PSNR was 70.94 dB for the Lena image using the hybrid wavelet-median method. It can be assumed that this new algorithm could also be applied to other agricultural images with good performance.