Motivated by the powerful computational capability in the emerging hardware, an applicable paradigm with its embedded lens calibrator is proposed. The proposed new paradigm for the relationship between the image provider and the image processor shows both the functional and economical merits. The paper first focuses on the developed of the embedded lens calibrator. An underlying support vector machine base regression (SVR) is hence employed as the key to achieve the goal. Based on the structural risk minimization, the SVR, employed as the calibration regressor, simultaneously minimize both the model complexity and empirical error, and create an estimator with a wide margin. The wide margin in regression represents a smooth approximation function for the lens calibration in which variances commonly existed in the CMOS camera modules can tolerably be eliminated. The variance tolerability achieves the calibration function a high robustness, and would conduct potentially the success of the proposed paradigm.