Hiding Data in the Blue Channel

Abstract

To enable the track and trace of pages from a color printer, data must be embedded and recoverable across the entire page regardless of the content on that page. This paper describes a solution that allows low visibility and high detectability of databearing “dot signatures”. An analysis of optimum dot-color to background-color combinations uncovers the serendipitous method of “Blue Clipping,” operating only on the blue component, exploiting low visual sensitivity to the blue channel. Blue Clipping is compared to the method of yellow dots, and the problem of dot gain for shadow dots is addressed. While initially designed for enterprise printers, this new technology is also adding functionality to 3D objects. It enables visually pleasing yet robustly recoverable data markings from a small region anywhere on a printed surface, without the need for costly special inks. Introduction How can we cover the surface of a document or 3D part with a data payload such that the data can be recovered from any subregion of the surface regardless of the underlying color, while minimizing impact on visual aesthetics, and without requiring costly invisible inks? This challenge was presented to us by customers in need of a solution for tracking sensitive documents, including trimmed portions of those documents. In this application a typical data payload could involve, for example, a user name, time stamp and location of the print. The problem of embedding information in documents has a long history of research. Since the 1980’s color laser printers have used a yellow microdot Machine Identification Code [1] on every page to enable identification of the device that printed the document, an approach that has been used to track leaked documents [2]. Anoto, Inc. provides technology [3] that covers an entire page with dots of the same size and shape where every dot is shifted from a nominal position as a form of encoding, but is not used to encode an arbitrary payload or survive recoverability in areas of non-white background. There are many 1D and 2D monochrome and color barcodes in wide use, but those patterns do not cover an entire document, do not recover reliably when overlaid on over content, and considerably change the visual appearance of the source document. Conventional watermarks will not work over dark solid areas, and, like barcodes, can be associated with minimal area requirements. Various efforts embed information in the halftone of continuous-tone images, and most apply to dispersed-dot dithering applications. A survey [4] of such techniques show that in many cases the entire image must be captured for payload recovery. A common approach is to use a form of error-diffusion. One such solution was developed at HP where the error diffusion algorithm used blocks of output pixel shapes [5]; this approach was built on earlier work, where the concept of a “visually significant barcode” (VSB) [6] was introduced. Other dispersed-dot dithering approaches include hiding data in the edges of the image [7], or toggling pairs of pixels [8]. Clustered-dot halftones have been used to carry information by creating asymmetric shapes in the clusters, such as ovals [9] and using shape orientation to encode a bit. Information can be embedded in clustered-dot screens by altering their phase and frequency [10], but the data payloads are limited to a small size. The technique proposed in this paper advocates a clustered-dot version of VSBs that embed data in a way that is visually pleasing by taking a conventional clustered-dot halftone and altering the positions of individual dot clusters [11]. The method is robust over a wide range of dot sizes across the full range of highlights and shadows and allows relatively large data payloads. We describe a solution using “dot signatures, ” featuring halftone-based VSBs derived from a fixed light gray image the size of the document. Figure 1 shows an example of a small portion of the resulting halftone in (a) and the VSB with data embedded in (b) at about 3x enlargement for illustration purposes. The data is arranged using a circular coding scheme [12] that allows full data recovery from a small portion of the pattern. We further describe how to render this pattern in a color document regardless of the background color and how to reliably recover the data. Figure 1. (a) Portion of a classical halftone of a fixed gray image, and (b) the resulting VSB carrying data. The Use of Yellow Dots One approach to consider is to overlay our dot signature as yellow dots because of its relatively low visibility over white space and recoverability when examining the blue channel of an RGB capture. As mentioned above, yellow dot Machine Identification Codes have been in use for almost four decades; this low visibility feature has also been used to hide data [13]. A test source document (a)