Browse Prior Art Database

Anti-aliasing de-mosaic processing by using a diamond-shape filter for digital imaging

IP.com Disclosure Number: IPCOM000010194D
Original Publication Date: 2002-Nov-04
Included in the Prior Art Database: 2002-Nov-04
Document File: 5 page(s) / 432K

Publishing Venue

Motorola

Related People

Sang Park: AUTHOR [+2]

Abstract

In digital color imaging pipeline, de-mosaic processing is one of the crucial steps especially in terms of digital image output quality. Authors demonstrate the de-mosaic processing in digital imaging can be functionally identical to anti-aliasing filtering in digital signal processing. Also, based on multidimensional signal processing theory, optimal filters for the de-mosaic processing are proposed, which are a rectangular-shape filter for the red and blue channel and a diamond-shape filter for the green channel.

This text was extracted from a Microsoft Word document.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 50% of the total text.

Anti-aliasing de-mosaic processing

by using a diamond-shape filter for digital imaging

Sang Park

David Hayner

Abstract

In digital color imaging pipeline, de-mosaic processing is one of the crucial steps especially in terms of digital image output quality. Authors demonstrate the de-mosaic processing in digital imaging can be functionally identical to anti-aliasing filtering in digital signal processing. Also, based on multidimensional signal processing theory, optimal filters for the de-mosaic processing are proposed, which are a rectangular-shape filter for the red and blue channel and a diamond-shape filter for the green channel.

 


Body

In digital imaging, a natural scene is taken by CCD or CMOS to be sampled as a digital pixel. But due to the current physical characteristics of CCD and CMOS, only one of RGB (red, green, and blue) information can be taken at a time. Since the green channel has more information than the blue or the red channel, a typical CCD or CMOS sensor forms a specific sampling pattern to capture the green channel information more than others, which is called Bayer pattern. Then collected Bayer-pattern samples need certain digital processing in order to have a full RGB information per pixel, which is called de-mosaic processing. Since the most following digital operations such as color transformation usually preserve the color information restored by the de-mosaic processing, it is very important how much color information can be restored in the de-mosaic processing.

The main purpose here is first showing the de-mosaic processing can functionally identical to interpolation, which is one of basic operations in multirate signal processing [1]. Then based on multidimensional signal processing theory, authors demonstrate that the performed sampling in the red and the blue channels is identical to rectangular sampling, and the sampling in the green channel to another sampling called quincunx sampling [2], respectively. Since any interpolation based on these sampling inevitably allows aliasing, the aliased frequency response need to be maximally removed, because it results in visual artifact in final outputs of digital imaging. Then it is well known that the rectangular-shape filtering and the diamond-shape filtering need to be used to minimize the aliased frequency response.

As seen in Figure 1, optical sampling between Figure 1 (a) and (b) can be considered as downsampling and upsampling based on the rectangular sampling. Then the de-mosaic processing between Figure (b) and (c) can be considered lowpass filtering for interpolation. Due to the rectangular sampling among Figure 1 (a), (b), and (c), only applying rectangular-...