Image denoising is an important part of the image pre-processing. It can affect results of image segmentation, edge detection, image recognition and image retrieval. Compared with the traditional denoising algorithms, such as median filtering and mean filtering, the denoising algorithm based on neural networks is much better. But because of the computational complexity of neural networks, the efficiency of neural-network-based denoising algorithm is low. Therefore, it is necessary to research the hardware implementation of neural-network-based denoising algorithm. This paper studies the hardware implementation of denoising algorithm based on ICM. In this paper, the main research work as follows:1. In this paper, the basic theory of neural networks and image denoising is learned and explored. And the hardware implementation of digital image processing is researched and analysised. These have laid a solid theoretical foundation for this research work.2. The design of image denoising system based on ICM is implemented by hardware programming technology. In this paper, the block diagram of hardware implementation is introduced, and is finished on the DE2-70development board. The important modules implemented include:image storage module, WY convolution module, ICM module, image denoising filter module and VGA display module.3. The design of image denoising system based on ICM is implemented by Niosâ…¡ technology. In this paper, the hardware and software implementations are both introduced, and are implemented on the DE2-70development board. The underlying hardware platform is finished in SOPC builder, and the software algorithm is finished in Nios â…¡ IDE.4. The experimental results based on hardware programming technology and NiosII technology are analyzed. The feasibility of the experiments is verified. The different level noises are added in experimental images. Then the noised images are processed by two methods. The experimental results are analysised and compared with Matlab simulation results. The results showed that the denoising effects are similar by FPGA hardware implementation and Matlab simulation. But the denoising speed of FPGA hardware implementation is faster than Matlab simulation. And the denoising speed based on hardware programming technology is faster than based on Niosâ…¡ technology. |