Abstract:Large-format full-frame transfer charge-coupled devices (CCD) feature a large array size, and their output images usually contain defective pixels. In addition, the brighter-fatter effect widely exists in CCD detectors. Both factors affect the temporal variance and gray-scale mean calculated from image data, resulting in inaccurate measurement of the device photon transfer curve and seriously reducing the accuracy of device parameter evaluation.To improve the evaluation accuracy of the nonlinearity of the photon transfer curve for large-format full-frame transfer CCD detectors, this paper studies image processing and correction methods. By partitioning, merging image data and eliminating defective pixels, the problem of reduced temporal variance of images caused by defective pixels is corrected, the correlation between adjacent pixels induced by the brighter-fatter effect is reduced, and the accuracy of the nonlinearity of the photon transfer curve and multi-channel gain of the device is improved.The results show that the photon transfer nonlinearity of the device is improved by an average of 97%, and the gains of each channel of the same device are corrected to the same gain range after image processing. The proposed algorithm enhances the measurement accuracy and efficiency of large-format full-frame transfer CCD devices, which has important application value and guiding significance for the research on detectors used in ultra-high-resolution scientific instruments.