Efficient companding algorithm suitable for color imaging

Details Efficient companding algorithm suitable for color imaging Efficient companding algorithm suitable for color imaging

2000-01-13

2002-02-26

Tran, Phuoc (Department: 2621)

Image analysis

Color image processing

C382S166000, C382S274000

Reexamination Certificate

active

06351555

ABSTRACT:

BACKGROUND OF THE INVENTION
1. Field of the Invention
The invention relates generally to the field of digital image processing. More specifically, the invention relates to methods and apparatus for image representation and enhancing digital image compression techniques.
2. Description of the Related Art
In digital imaging, which is the process of capturing and representing into digital form a real image, there is a tradeoff between the size and accuracy of the digital image. Typically, as the size of an image (when digitally represented) is increased, the resultant quality of the digital image (when compared to the original) is also increased. One particular step in the imaging process where this is particularly true when a captured image undergoes “companding.” Companding is the procedure by which a higher resolution pixel is reduced to a lower resolution. The typical image sensor, the focal device which initially captures a light or color intensity of an image, generates pixels of 10 or more bits based on the analog-to-digital conversion characteristics. Thus, for example, when a digital still camera takes a picture, the image is first represented by the sensing elements as a group of color/intensity values which can have 2
10
or 1024 possible values. Pixels from an image sensor that compose a captured image however, are not optimized for most image processing/compression algorithms which are designed for pixel values in multiples of 8 bits. This is due primarily to the fact that image processing has traditionally been performed on computer systems where the unit of a “byte” (8 bits) served as a basis of computer system architectures. Thus, a vast majority of image compression schemes and techniques, as well as image representation methods, are designed around the 8-bit (byte) model. To take advantage of these pre-existing techniques and to interface digital cameras with computer systems, it would thus be advantageous to reduce all pixels of 10-bit resolution (or any resolution greater than and not a multiple of 8 bits) to pixels of 8 bit resolution.
While doing so, it would also be desirable to preserve key image information such as edges, shading and color. The crude method for companding involved a mere truncation of the 2 least significant bits. This equates to dividing the pixel value by four. Such a simple linear transformation does not take into account the statistical properties of the image, i.e., how pixel values are clustered or distributed throughout the image. For instance, if an image contains a pixel value that repeatedly occurs, and then at a particular location, changes slightly, this information should not be lost. Assume for example a solid color region of an image which fades into a different color. The boundary pixels between the solid color and the faded portion should be appropriately represented. For instance, if the solid color region has values of say 514 and the boundary pixel a value of 513, the traditional companding technique would educe these pixels both to a value of 128. Thus, the edge information represented by the boundary pixel value 513 between the faded region and the solid color region is lost and the edge will disappear in the companded image.
The end result visually to the human eye of any companding is a change in the contrast of each pixel from bright to dark. If the companding process is performed as a straightforward linear mapping as discussed above the change in contrast will reduce detail information of the image, and thus, lower the image quality. However, if companding can be achieved in a manner more indicative of the characteristics of the image, then the resultant change in contrast will also preserve the essential image features or characteristics.
It would be desirable to compand in an intelligent manner to preserve more details and colors of an image by taking into account the characteristics of the image itself. Further, it would be advantageous to compand in a manner that is suited to implementation in VLSI (Very Large Scale Integration) circuits such as those found in low cost imaging devices.
SUMMARY OF THE INVENTION
What is disclosed is a method including determining a set of factors related to the color characteristics of a captured image and then companding the captured image into a companded image having a lower bit-per-pixel resolution, the companding dependent upon that set of factors.


REFERENCES:
patent: 5181105 (1993-01-01), Udagawa et al.
patent: 5708729 (1998-01-01), Adams et al.
patent: 5790705 (1998-08-01), Anderson et al.
patent: 0516083 (1992-12-01), None

Affiliated with

Also associated with

Rating

Efficient companding algorithm suitable for color imaging does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Efficient companding algorithm suitable for color imaging, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Efficient companding algorithm suitable for color imaging will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-2983897