A fixed length code (FLC)-based image compression method and a device thereof. The method is utilized to compress a block containing plural pixels and includes the following steps: determining a first representative pixel, a second representative pixel and a third representative pixel from the pixels according to pixel values of the pixels, the three representative pixels being noncollinear in a color space to which the pixels correspond; generating plural first interpolated pixels by interpolation according to the first representative pixel and the third representative pixel; generating plural second interpolated pixels by interpolation according to the second representative pixel and the third representative pixel; and generating an index value for each pixel according to the three representative pixels, the first interpolated pixels and the second interpolated pixels.