Digital Image and Video Processing Using Subpixel Rendering.

摘要

Subpixel rendering techniques originate from the problem of monochromatic font rendering on LCDs. It takes advantage of the fact that a single pixel on a color LCD display consists of several primary colors, typically three colored stripes (subpixels) ordered red, green, and blue (RGB). Researchers found that, by controlling the subpixel values of neighboring pixels, it is possible to micro-shift the apparent position of a line to gives greater details of text.;In this thesis, we address the problem of color image and video processing using subpixel rendering techniques to achieve superior sharpness for small LCD displays by controlling individual subpixels rather than pixels. However, the increased luminance resolution often comes at the price of chrominance distortion. A major challenge is to suppress color fringing artifacts while maintaining sharpness.;First, we discuss subpixel rendering for down-sampling problem, such a problem exists when a high resolution image or video is to be displayed on low resolution display terminals (i.e., Mobile). We start by formulating subpixel-based down-sampling as different optimization problems based on different reconstruction models in spatial domain: MMDE (Min-Max Directional Error) and MMSESD (MMSE for subpixel-based down-sampling). Simulation results show our proposed subpixel-based methods can give much sharper images compared with the conventional pixel-based methods, without noticeable color fringing artifacts.;To better understand what happens in subpixel-based algorithm, we further propose novel frequency domain analysis approach to explain why it is possible to achieve a higher apparent resolution using subpixel techniques. Our theoretical analysis shows that the cut-off frequency of the low-pass filter for subpixel-based decimation can be effectively extended beyond the Nyquist frequency using novel anti-aliasing filters.;We also investigate the problem of low bit-rate JPEG compression. Since JPEG introduces severe blocking artifacts under low bit-rate, researchers have proposed a down-sampled image when compressed and later interpolated scheme, which provides superior performance than high resolution image compressed directly. However, pixel-based down-sampling loses high frequency details, resulting in blurring reconstructed image. We thus proposed subpixel-based low bit-rate JPEG compression algorithm without changing decoding system, which can provide about 3dB higher PSNR than pixel-based low bit-rate JPEG compression scheme, under the same bit-rates.;Finally, we address the problem of displaying high resolution one-color Bayer image on low-resolution LCD screen of portable devices, which requires demosaicking followed by down-sampling. However, these two steps require high computational complexity, and the color artifacts introduced in demosaicking will be magnified in down-sampling. We thus propose joint demosaicking and subpixel-based down-sampling algorithm by directly performing subpixel-based down-sampling in Bayer domain without demosaicking, resulting in greatly reduced computational complexity and sharper down-sampled images.