In this paper, a robust yet simple and efficient method is proposed for a 2D image encryption. The purpose of this work is to plug the gap related to blind decryption and lack of randomness in optical transform domain which limits its application in encryption of data. The novelty of this work is in the method of generating multiple transform orders. Two piecewise linear chaotic maps (PWLCM) are used to generate multiple transform orders along both the dimensions. A fractional Fourier transform is obtained with the generated multiple order vectors. The transform domain data is scrambled with yet another integrated chaotic map to further enhance the security. In the proposed work, the chaotic maps are selected, keeping in view, limitations related to 1D chaotic maps. The PWLCM and an integrated chaotic map are robust due to absence of periodic windows that exist in 1D chaotic maps. A reality preserving algorithm is used to overcome the limitation of dealing with complex transform coefficients for storage and transmission purposes thereby making it suitable for fast and real time processing. The results and experimental analysis prove that proposed scheme is highly random in encrypted domain and it possesses excellent sensitivity. The entropy in encrypted domain is considerably higher than other similar state-of-art schemes. The scheme is robust against classical attacks, noise attack, differential attacks and entropy attack. The present work is confined to 2D images, however the proposed scheme can be further extended to multispectral images.