Edge detection can be used both in spatial and frequency infomation. However, the calculation complexity detection in spatial domain is much greater than that in frequency domain. To compressed image, it is more obvious because the processing should include two parts: frequency to spatial and spatial to frequency. So it cannot satisfy the dealing-in-time for video. Discrete cosine transform (DCT) is widely used in image processing and compression for its characteristics of low calculation complexity, orthogonality and fixed primary function. Based on the relation between DCT coefficients and spatial information, a fast edge detection method is proposed. In the method, main edge direction is detected by using only two low frequency coefficients F(1,0) and F(0,1).