Dissertation > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device

A New Image Fusion Algorithm Based on Wavelet Transform and PCNN

Author HuYiLing
Tutor ZouBeiZuo
School Central South University
Course Computer Science and Technology
Keywords wavelet transform image fusion pulse coupled neural networks
CLC TP391.41
Type Master's thesis
Year 2010
Downloads 127
Quotes 1
Download Dissertation

With the revolution of new technology, Our world entry into the full information era. A new theory and technique was developed which is really effective in the multi-source information fusion. Fusion as a visual information fusion, has become a widespread concern for most people around the world.In our article, we combines wavelet transform with pulse coupled neural networks in image fusion.The main research contents include the following three parts.①We compared and discussed some traditional pixel-level image fusion algorithms just like simple image fusion, the fusion method based on pyramid decomposition and the fusion method based on wavelet transform. The widely used wavelet transform is described in detail. We also explained how to use the wavelet decomposition for image fusion process, and specifically describes the fast wavelet transform algorithm. Compared with traditional wavelet transform, the fast wavelet transform have the better time-frequency characteristics.②Pulse coupled neural network model was introduced in the paper, we analysis the parameters in the model, and find in traditional methods the connection coefficient is a fixed value. In order to get rid of the limitation, we present using image spatial frequency after wavelet transform to dynamically adjust the connection coefficient.③Aiming at the limitation of traditional wavelet transform, our paper propose a new image fusion novel that after decomposing the image with fast wavelet transform, we use pulse coupled neural networks to choose fusion rule instead of artificial selection. The key point is in allusion to the feature of high frequency and low frequency, using different rules, respectively. The option on high frequency part is to use pulse coupled neural networks. As carrying little information of image, and effecting little on the final fusion. On the low frequency part, we still choose the traditional weighted average method.Through performing experiment to two sets of images from different sensors, and analysising the subjective observations and objective evaluation standard about the result of fusion, shows that this algorithm can preserve more useful information from original images effectively, and enhance the quality of the fused image.

Related Dissertations
More Dissertations