The invention relates to an image compression method. Computer equipment divides an image to be compressed into a plurality of compression units; wherein each compression unit comprises at least two sub-blocks; then, on the basis of a preset compressed data format, compressed data of each compression unit is obtained; wherein the compressed data comprises a reference pixel value of the compressionunit, first pixel associated data, second pixel associated data, a sub-block position where the reference pixel value of the compression unit is located, and a pixel position where the reference pixel value of each sub-block is located; wherein the first pixel associated data is used for representing the compression unit and the pixel relationship of each sub-block in the compression unit, and the second pixel associated data is used for representing the pixel relationship in each sub-block. By adoption of the method, the compression ratio of compressing the to-be-compressed image can be improved, the bandwidth utilization efficiency is improved, and the time delay of compressed data reading is reduced; and information distortion caused by compression of the to-be-compressed image is avoided.