Disclosed herein are techniques for pre-processing image data for compression, e.g., image data that represents burn-in statistics for a display device. The techniques can involve receiving the image data, where the image data comprises a plurality of pixels, and each pixel of the plurality of pixels comprises at least two sub-pixel values. Next, each pixel of the plurality of pixels is quantized to produce a plurality of modified pixels. Subsequently, a series of operations are performed against each modified pixel of the plurality of modified pixels, including (1) applying an invertible transformation against the modified pixel, (2) applying a predictive coding against the modified pixel, and (3) applying an encoding of the modified pixel into a buffer as a data stream. The buffer is then compressed (as the modified pixels are serially encoded into the buffer) to produce compressed outputs that are joined together to produce a compressed image.