eduzhai > Applied Sciences > Engineering >

Joint Demosaicking / Rectification of Fisheye Camera Images using Multi-color Graph Laplacian Regularization

  • Save

... pages left unread,continue reading

Document pages: 5 pages

Abstract: To compose a 360 image from a rig with multiple fisheye cameras, aconventional processing pipeline first performs demosaicking on each fisheyecamera s Bayer-patterned grid, then translates demosaicked pixels from thecamera grid to a rectified image grid---thus performing two image interpolationsteps in sequence. Hence interpolation errors can accumulate, and acquisitionnoise in the captured pixels can pollute neighbors in two consecutiveprocessing stages. In this paper, we propose a joint processing framework thatperforms demosaicking and grid-to-grid mapping simultaneously---thus limitingnoise pollution to one interpolation. Specifically, we first obtain a reversemapping function from a regular on-grid location in the rectified image to anirregular off-grid location in the camera s Bayer-patterned image. For eachpair of adjacent pixels in the rectified grid, we estimate its gradient usingthe pair s neighboring pixel gradients in three colors in the Bayer-patternedgrid. We construct a similarity graph based on the estimated gradients, andinterpolate pixels in the rectified grid directly via graph Laplacianregularization (GLR). Experiments show that our joint method outperformsseveral competing local methods that execute demosaicking and rectification insequence, by up to 0.52 dB in PSNR and 0.086 in SSIM on the publicly availabledataset, and by up to 5.53dB in PSNR and 0.411 in SSIM on the in-houseconstructed dataset.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...