Research And Hardware Design Of Low Illumination Image De-noising Algorithm

With the development of multimedia technology, image processing has penetrated all aspects of life, and image de-noising is an essential step in image processing. In some special applications under the condition that such as lack of illumination or uneven illumination, the collecting images constrain high amounts of noise, vague edge and even some information can not be identified, these seriously affected the subsequent use of image, so image de-noising of this type is important.Image de-noising can restrain noise and get the effective edge information. There have emerged quite a few classical methods and theories in the research of image de-noising, but these algorithms are usually focus on the image which under ideal illumination condition. According to the characteristics of low illumination image, this paper combined edge detection and image de-noising, and then proposed a mixed local threshold de-noising algorithm based on improved canny edge detection; on this basis, the paper proposed a mixed de-noising algorithm based on pixel correlation for the test images; Finally, according to the requirements of real-time applications, the paper improved the traditional median filtering algorithm, and designed spatial mixed filtering algorithm on FPGA.The main contents are as follows:(1) Analyze the principles of traditional canny algorithm, Then this paper propose an improved canny edge detection algorithm based on it.(2) Study the key factors which affect wavelet threshold de-noising, analyze the principles and characteristics of Bayes threshold and then propose an improved adaptive partial threshold algorithm. According to the noise diversity, the paper de-noises by local threshold de-noising mixed algorithm which based on the improved canny edge detection.(3) Introduce the character of text image. A spatial de-noising algorithm based on the pixel correlation is proposed by analysis of the deficiency of stroke fracture and false fracture when the conditional spatial filter is been used, and then de-noised the low illumination text image by combining of wavelet threshold method.(4) Complete the hardware design of image real-time de-noising algorithm. Considering the real-time and the complexity of algorithm, the paper improves the hardware design of median filter algorithm and presents a mixed algorithm based on the mean filter and weighted average filter.The results show that the low illumination images de-noising algorithm is suitable for a variety of applications. The spatial hybrid filter designed on FPGA has strong versatility and may be applicable for many places such as real-time supervisor, visual presenter, infrared measurement system and other occasions.