HP Color Recovery technology - for low-cost color image display - Technical

Hewlett-Packard Journal, April, 1995 by Anthony C. Barkans

The Filter Process

In the example given earlier a red color component represented by the binary value 01011000 (2.75 in decimal) was used to illustrate simple dithering. For this example we used a 2 x 2 dither region in which the end result of the dither was that 3/4 of the pixels stored in the frame buffer were set to 3 (011) and 1/4 of the pixels were set to 2 (010). It is easy to see that if we average the four pixels in the 2 x 2 region we will recover the original color. This can be done as follows:

([value_1 x number_set_to_value_1] [value_2 x number_set_to_value_2])/total_number_pixels

Using the example data we obtain: ([3 x 3] [2 x 1])/4 = 2.75.

This averaging works very well in regions of constant color, such as the sky behind the jet plane in Fig. 1. However, there is one fundamental issue that must be addressed for HP Color Recovery to be viable and that is how to handle edges in the image. If edges are not accounted for then the resultant image will blur. The two-dimensional representations of an area of a display screen shown in Fig. 4 are used to illustrate the problem of edge detection and the way the problem is addressed in HP Color Recovery.

[ILLUSTRATION OMITTED]

As in Fig. 2, each box represents a pixel location on the display screen. In Fig. 4a the numbers represent the original true color data for one of the color components (e.g., red) in a 24-bit per pixel system. Fig. 4b shows the same region after simple dithering has been applied. Fig. 4c shows the pixel values after the application of HP Color Recovery. Fig. 4c pixel values represent the color data that would be displayed on the computer screen.

[ILLUSTRATION OMITTED]

Region A in each of these figures is an area of constant color, whereas region B encompasses an edge. For illustration purposes, the dither region is again assumed to be 2 x 2 pixels.

The dithered color data shown in Fig. 4b is derived from the original color data shown in Fig. 4a and from using the simple dithering technique described in connection with Table I. The data shown in Fig. 4b is what would be stored in the frame buffer and displayed in a typical dithered system (e.g., Fig. 1b).

[ILLUSTRATION OMITTED]

When it is time to display Pix_1 the data for the four pixles shown as Region A in Fig. 4b would be sent to the filter. The data stored in the region would be summed and then divided by the number of pixels in the region. The sum of the pixels in Region A is 11 and 11/4 = 2.75. Thus, the output of the filter when evaluating Pix_1 would be 2.75. This output value would be displayed on the computer display at Pix_1's location. Note that the output of the filter is the exact value of the original data at that point in Fig. 4a.

[ILLUSTRATION OMITTED]

The next pixel along the scan line to be evaluated is Pix_2. The filter region for evaluating Pix_2 would include the two rightmost pixels of region A and the two leftmost pixels of region B (see Fig. 4b). Applying the filter operation for Pix_2 again results in the output value matching the value at that location in Fig. 4a (2.75).