Still Image Reconstruction And Compression For CCD Samples Based On Bayer Pattern

This dissertation mainly discusses Bayer data compression and CFA demosaicing. CFA demosaising is necessary before fully RGB color image. Advanced CFA demosaicing algorithms are some kind of property of camera manufacture. Bayer data compression is quite recent issue. It is significantly to develop research in these fields, since scarcely any people care about the CFA demosaicing and even few about the Bayer data compression.In a single-chip digital color-imaging sensor, a color filter array (CFA) is used to obtain sampled spectral components (red, green and blue) in an interleaved fashion. Color demosaicing is the process of interpolating these regularly spaced sampled values into the dense pixel maps for each spectral component.CFA demosiacing method is relatively developed. A wide discussion about most of demosaising algorithms today is present. Several simple and fast examples are given. Whatever algorithm spatial smoothness and color correlation are two mainly factors which effect reconstructed image quality. In this paper two improved CFA demosaicing methods are proposed. The first one isbased on decomposition and uses inter-channel correlation effectively in an alternating-projections scheme. In the second method we present techniques for interpolating the color images in the YUV color space. The resulting interpolated images could be directly used in the DCT based JPEG compression scheme.This paper describes two compression method suitable addressed for Bayer data compression. In the first proposed approach, the input images acquired by the sensor in CFA format is compressed with JPEG luminance quantizing table. Since the quantizer performs a fine quantization in luminance quantizing table, information losses is reduced. In the second algorithm we introduces a new, efficient coding method to compress the Bayer Pattern. It is based on a predictive schema followed by a Vector Quantization(VQ)technique. Simulations have demonstrated that the proposed scheme allows a visually lossless compression of Bayer pattern images with low memory cost.