The WH LFBC PDF shows white ‘halos’ around the text and when the text is ‘removed’ by turning off the layer, the background shows a whitish area wit hints of colors. This has been referred to as the x-ray effect.

The x-ray effect can be understood by realizing that the whole background is encoded in JPEG which means color will leak into these ‘white areas’. Furthermore, some MRC compression methods use a concept of ‘data filling’ to reduce the gradients between the background and the ‘holes’ created by the separation of the text as this reduces JPEG artifacts (ringing)

Let’s try to understand the MRC workflow better

1. The document is scanned by the WorkCentre (maximum of 600 DPI). Let’s assume that it is scanned at its maximum.

2. The document then segmented into fore and background by analyzing 16×16 blocks

Here there are two reasonable pathways:

3a. The Background is encoded in JPEG and subsequently sub-sampled to 1/4 of the resolution (150 PPI)


3b. The Background is sub-sampled to 150 PPI and then encoded in JPEG

4. The foreground objects are sub-sampled to 300 PPI.

This is supported by the observation that the JPEG is stored as a 150 PPI image and the foregrounds at 300 DPI, even though they are rendered on a 72 DPI grid in the PDF. MRC usually sub-samples the foregrounds less than the background as they contain text that is far more sensitive to sub-sampling. The JPEG encoding is set to a quality of about 50, which is a very high compression, and which leads to visible JPEG artifacts.

Also note that there are different ways to sub-sample as four pixels or 16 pixels (JPEG) become one pixel in the new image. How to determine the new values can affect the analysis. Typically, some averaging is used to smooth the transitions.

Things get worse for JPEG as the different channels Y, Cb and Cr are themselves subsampled. Typically the Y channel remains the same and the Cb and Cr channels are sub-sampled by a factor of two, creating a MCU (minimum coded unit of 16×16). This sub-sampling scheme is know as 4:2:0. As to why JPEG uses the YCbCr color scheme and why it uses 4:2:0 compression, it is helpful to read up on JPEG. The short explanation is that our eyes are very sensitive to variations in luminance (brightness) and changes in green, Cr and Cb (or U an V) encode red and blue values and thus they affect green least. JPEG compression is all about human vision.



One thought on “Halos

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s