An edge enhancement method and apparatus for Bayer images are provided. The apparatus stores the raw Bayer image data into a line buffer. After having directly read the raw Bayer image data from the buffer, the apparatus computes luminance characteristic values, analyzes the edge characteristic and then outputs the edge characteristic values. According to the edge characteristic values, the edges in the raw Bayer image data are enhanced. The apparatus simplifies the processing complexity of the rear-stage circuit by directly enhancing the edges in the Bayer image, and is easily combined with the post image processing mechanism. Conventionally, the raw Bayer image is converted to the RGB image or the YUV image and then the edge enhancement operation is performed. The invention reduces the computing volume and memory size. Also, performing the edge enhancement operation in advance helps to identify the interpolation direction and ease the annoying noise.