Sparse decomposition is one of the core issue of compressive sensing ghost image.At this stage,traditional methods still have the problems of poor sparsity and low reconstruction accuracy,such as discrete fourier transform and discrete cosine transform.In order to solve these problems,joint orthogonal bases transform is proposed to optimize ghost imaging.First,introduce the principle of compressive sensing ghost imaging and point out that sparsity is related to the minimum sample data required for imaging.Then,analyze the development and principle of joint orthogonal bases in detail and find out it can use less nonzero coefficients to reach the same identification effect as other methods.So,joint orthogonal bases transform is able to provide the sparsest representation.Finally,the experimental setup is built in order to verify simulation results.Experimental results indicate that the PSNR of joint orthogonal bases is much higher than traditional methods by using same sample data in compressive sensing ghost image.Therefore,joint orthogonal bases transform can realize better imaging quality under less sample data,which can satisfy the system requirements of convenience and rapid speed in ghost image.