Color filter arrays revisited — Evaluation of Bayer pattern interpolation for industrial applications

A Recursive Edge Detector For Color Filter Array Image

“…Starting from Bayer [2] observation: green channel can be assimilated to luminance, red and blue channels to chrominance. SEDD (Smoothed component Estimation and Deriche Derivative) computes a gradient by estimating bilinear interpolations [7]. Here, we propose to decompose the CFA image into two channelsÏ G andÏ R,B which both contain half of all pixels arranged in a quincunx lattice (see Figs.…”

“…( 1) and (2). As diagonal pixels are separated by a distance of √ 2 instead of 1 in usual directions (the gap between each pixel of a color channel in a CFA image), the parameter α must be modified so α = √ 2α, [7]. Each channelÏ G andÏ R,B is diagonally smoothed using eq.…”

“…1 shows the computational complexity of those algorithms. The complexity is evaluated as the number of required elementary operations per pixel: addition, subtraction, multiplication and division, each one being weighted similarly, as detailed in [7]. The expressions of the recursive Deriche (recursive of order 2) equations require seven additions and eight multiplications per pixel for the smoothing filter, seven additions and nine multiplications per pixel for the derivative filter [10].…”

“…Aberkane et al [7] showed that directly edge detection in those images is possible with a very close or even better precision than segmentation after demosaicking. They introduced two approaches to use the raw data of CFA images for edge detection: (i) Deriche-based Luminance (DL) technique estimates partial derivative images of luminance thanks to Deriche recursive filters [8], so that a gradient image is computed, (ii) Deriche-based Color (DC) derivative method also uses Deriche filters to compute partial derivatives of red, green and blue channels so that a structure tensor is applied [9].…”

“…Despite the best results offered by DL, methods computing luminance derivatives from Bayer CFA were not studied deeply enough. Aberkane et al [7] showed the failure to detect isoluminant edges. Moreover, a raw image of luminance could be directly captured with a monochrome camera.…”