To solve the problem of low and non-uniform resolution in catadioptri c omnidirectional imaging,the theory of compressive sensing (CS) is applied to research the catadioptric omnidirectional imaging sys tems. The resolution distribution of catadioptric omnidirectional imaging systems is analyzed based on the unit sphere model.Acco rding to the relation between mean square error (MSE) and sparsity,measurement number,a non-uniform measurement matrix,which is ba sed on the distribution of the imaging system′s resolution,is designed in this paper.Based on the designed measurement matrix, it allocates more sensing resources to inner parts but fewer to outer parts of the catadioptric omnidirectional image.This scheme take s the resolution of omnidirectional image into account, which is increased from the center to the periphery.The non-uniform compressed samples of the observed scene are captured from the ray light which is reflected from the mirror.The linear Bregman iteration is employ ed as the reconstruction algorithm to obtain the high and uniform resolution image.The algorithm is tested on synthetic and realistic ima ges.Experimental results show that the proposed method is feasible and effective.The recovered image has higher and much more uniform resolution than that reconstructed by traditional method.