For the non-uniformity of polarizer array caused by the difference of the lithography accuracy and the transmission of each pixel polarizer in a pixel polarized camera, a least squares based super-pixel calibration method with angle constraint was proposed. A mathematical model of calibration optimization algorithm with angle constraint was established, meanwhile, scalar gain and dark compensation parameters were added in the model. Finally, an integrating sphere calibration system by taking a micro-polarizer as the core was established, simulation experiment and imaging experiment were performed. The experimental results show that the proposed method in this paper corrects effectively the amplitude unevenness of the super-pixels while correcting the degree of polarization and non-uniformity as compared with the least squares method, and the accuracy of the reconstruction perspective has increased by 10%. After interpolation processing, the edge performance is more obvious, and the noise generated during the interpolation process effectively is reduced and the image quality is improved.