Comments:

Lossless JPEG is an extension to the normal JPEG standard. One of the main algorithmic differences between the two is that the lossless JPEG does not apply a Discrete Cosine Transform. Rather, it uses a Predictive scheme. For each pixel, the values of one or several neighboring pixels are added to the value of the original pixel and then subtracted from the value of the original pixel. This method yields smaller values that require fewer bits per pixel to store.

ImageGear allows you to set the number of neighboring pixels to use in calculating the "predictor value".