| Summary: | 碩士 === 國立中興大學 === 電機工程學系所 === 101 === To reduce costs and hardware sizes, digital cameras have recently been developed to possess image sensors for capturing images. A commonly used color filter array (CFA) that covers the image sensors is the Bayer CFA. CFAs direct light to pixels, and each pixel in the image sensor captures the intensity of red, green, or blue light. The light captured by the Bayer CFA is then reconstructed to create full-color images. This reconstruction process is known as demosaicking or CFA interpolation.
This study proposes a highly efficient modified algorithm based on the primary-consistent soft-decision algorithm. This method is characterized according to two features: (1) soft-decision, typical color difference planes (R-G or B-G) are transformed to planes in chrominance spaces (YUV or UV); and (2) instead of using a common median filter, this algorithm employs a false-color reduction process to significantly reduce false color areas that are caused by poor decision. In addition, the experimental results indicated that full-color images, which are processed twice using the proposed method, are visually similar to raw images, and image quality can be enhanced.
In this paper, Section 1 introduces commonly adopted demosaicking algorithms and presents a comparison of their advantages and disadvantages.
Section 2 describes the proposed algorithm. The transformation from color difference planes to planes in chrominance spaces and the false-color reduction process were used to reduce false-color areas that could not be determined by a soft decision. Light is processed twice using these methods to export full-color images that are similar to the raw images.
Section 3 shows the results in which the proposed algorithms visually and statistically improve the image output and quality.
Section 4 provides the evaluation of image errors caused by the image interpolation-restoration method of various algorithms. A small error implies that the quality of the restored images is similar to the raw images. Section 5 offers suggestions for the experiments conducted in this study.
|
|---|
