Haar wavelet is one of the best mathematical tools in image cryptography and analysis. Because of the specific structure, this wavelet has the ability which is combined with other mathematical tools such as chaotic maps. The rational order chaotic maps are one of clusters of chaotic maps which their deterministic behaviors have high sensitivity. In this paper, we propose a novel method of gradient Haar wavelet transform for image encryption. This method use linearity properties of the scaling function of the gradient Haar wavelet and deterministic behaviors of rational order chaotic maps in order to generate encrypted images with high security factor. The security of the encrypted images is evaluated by the key space analysis, the correlation coefficient analysis, and differential attack. The method could be used in other fields such as image and signal processing.