By summarizing various algorithms of blind pixels detection and compensation, we can divide blind pixels into four types for their different respondent appearance, which are dead pixels, dark current pixels, noisy pixels and blinking pixels. Different algorithms with different thresholds are proposed to detect different types of blind pixels. Compared with the traditional algorithms, the algorithm takes dark current pixels and blinking pixels into account and has high precision of detection. Based on the pertinence between the adjacent pixels, a compensation algorithm is advanced to detect congregated blind pixels and replace the blind pixels on the edge of image. Experiments show that the algorithm is practical and acceptable.