Research On Noise Reducation And Color Interpolation Algorithms Based On Bayer Pattern Images For Digital Camera

As the improvement of image sensor technology and digital image process theory, digital camera has been widely used today. This dissertation researches into the noise reduction and color interpolation algorithms for digital camera. The noise reduction is the basic guarantee of the image quality, and the color interpolation is the key algorithm for the reconstruction of full color image from raw data. Both of them play important roles in a digital camera.Considering that salt-pepper noises and white Gaussian noises exist commonly in a digital camera, a new adaptive noise reduction algorithm is proposed in this paper to reduce these two noises. Firstly, a parameter-estimation method based on the characteristic of Gaussian noise is proposed to obtain the standard-deviation. The result of estimation is used for the salt-pepper noises detection, and the pixels affected by salt-pepper noise will be processed by median-filter to suppress noise. And then, an algorithm based on linear-fitting of neighboring similar pixels is adopted to reduce white Gaussian noise. Since the spatial correlation of image is considered, this algorithm can reduce the noise efficiently while preserving the details of image.Considering the character of the distribution of Bayer pattern image, a color interpolation algorithm based on color difference space and edge detection technique is proposed. A new method which use color correlation of image to detect edge is utilized to improve the precision of edge detection. Moreover, the interpolations of all three color components are processed in color difference space. With this interpolation algorithm, we can reconstruct the full color images without making the image detail fuzzy, and the false color distortion will be suppressed efficiently too.