The invention provides a background suppression algorithm for steel coil end face defect detection, which comprises the following steps of: inputting a local grey-scale map of a steel coil end face, and carrying out two-dimensional discrete time-frequency conversion on a local grey-scale map two-dimensional array f (x, y); equally dividing The spectrogram two-dimensional array F (u, v) into four quadrants; translating the diagonal quadrant of the two-dimensional array F (u, v) matrix of the spectrogram for position exchange to obtain a translated spectrogram F _ shift (u, v); converting the translation spectrogram into an amplitude spectrogram, and splitting the amplitude spectrogram two-dimensional array M (u, v) into a low-frequency component spectrogram two-dimensional array M _ low (u,v) and other component spectrogram two-dimensional arrays M _ else (u, v); Setting pixel points corresponding to the background frequency spectrum direction in the translation frequency spectrum two-dimensional array F _ shift (u, v) to be zero so as to obtain a new frequency spectrum two-dimensional array F _ field (u, v); And carrying out two-dimensional discrete time-frequency inverse transformation on the two-dimensional array F _ field (u, v) of the new spectrogram to obtain a local grayscale image f '(x, y) after background suppression. The method can be used for carrying out backgroundsuppression on the end surface of the steel coil, and compared with the existing filtering algorithm, the effect is greatly improved.