A quantum-inspired noise reduction method based on noise feature codebook

Noise reduction is a basic problem in image processing, binary image contains less information than color images, which means it´s more difficult to reduce noise in binary image. The most used noise reduction methods are mathematic morphology and median filter, but they face a common problem that the fine structure information in the image was destroyed in the process of noise reduction. Inspired by the quantum theory, we express an image with quantum system first, after that, a new distance measurement between quantum states was proposed, then we advance a concept, noise feature codebook, to describe the local feature of each pixels in the image. The distribution of the codebook in images with difference noise density was studied, the relationship of the distribution and the noise density was built up by training of masses of images. With the regularity, a noise reduction method was proposed, which choose optimal noise judging criterion according to noise density of the image. Experiment showed that proposed method gain higher SNR and better visual quality than tradition noise reduction methods in most situation.