Middle-wave HgCdTe detectors are the most commonly used high-performance cooled infrared focal plane array (IRFPA) detectors, which are widely used in many fields. The middle-wave mercury cadmium telluride detector has a type of isolated pixel in actual use. It has a slightly lower voltage level than the surrounding pixels, but it does not meet the blind pixel judgment standard of the national standard, which will affect the use in photoelectric systems. We named them sub bad pixels. Relying on human eyes to judge such pixels not only takes a lot of time, but also different people have different judgment standards for sub bad pixels. Aiming at the difficulty of judging the sub bad pixels mentioned above, through analyzing the judgment process of the human eyes on the sub bad pixels, and using the computer to simulate the human judgment, the judgment standard of the sub bad pixels is quantified, and the automatic identification and statistical functions are realized. Compared with the original human eye judgment method, the accuracy and efficiency of recognition are greatly improved.