This paper proposes an image steganography for hiding secret images. The objective of the proposed work is to increase the embedding capacity while ensuring the security of the image. This has been addressed using two levels of processing. First is the application of Genetic Algorithm (GA) as ‘before embedding hiding technique’ which tries to identify suitable places in cover image where embedding of secret image will not lead to much distortion in the image. Second process is improvement using Discrete Cosine Transform (DCT) and its wavelet (DCWT) to gain the advantage of their energy compaction property. The secret image is embedded in the lower energy transformed regions of the cover image. The paper makes a comparison between DCT and DCWT (with and without using GA). The experimental results are evaluated using various performance evaluation parameters such as Mean Square Error (MSE), Peak Signal-to-Noise Ratio (PSNR) and Correlation. The overall comparison proves that both DCT and DCWT perform better when combined with genetic algorithm. This work also proves that the proposed two level processing contributes in improvement of both security as well as embedding capacity for the image steganography.