Multiple watermarking techniques for images is receiving more attention in recent years for its wide variety of applications in different fields such as piracy of digital data, and copyrights protection. Current approaches rely on adding many watermarks in different bands or channels utilising scaling factor, and embedding locations that are mainly defined by experts. This brought many challenges in achieving equilibrium between security, robustness, and quality. Aiming to fill the gap of determining stationary scaling factor and embedding locations, the work proposed in this paper adapts genetic algorithm to find the optimality of these issues; that will enhance both of watermarking capacity and imperceptibility in colour images. To increase security, the suggested model encodes the watermarks using 2-D Walsh transform (WHT). Furthermore, Singular Value Decomposition (SVD) as a dimension reduction tool is utilised to factorise Wavelet coefficients to determine the most salient embedding locations. Experimental results show that the proposed model is more robust against common image manipulation attacks in terms of PSNR and NCC.