QQ客服 官方微博 反馈留言

浏览/检索结果: 共3条，第1-3条 帮助 限定条件 专题：长春光学精密机械与物理研究所    第一署名单位    第一作者单位    通讯作者单位     已选(0)清除 条数/页：5101520253035404550556065707580859095100   排序方式：请选择作者升序作者降序题名升序题名降序发表日期升序发表日期降序提交时间升序提交时间降序 Multi-focus image fusion algorithm based on adaptive PCNN and wavelet transform (EI CONFERENCE) 会议论文 International Symposium on Photoelectronic Detection and Imaging 2011: Advances in Imaging Detectors and Applications, May 24, 2011 - May 26, 2011, Beijing, China Wu Z.-G.; Wang M.-J.; Han G.-L. 收藏  |  浏览/下载：34/0  |  提交时间：2013/03/25 Being an efficient method of information fusion  image fusion has been used in many fields such as machine vision  medical diagnosis  military applications and remote sensing.In this paper  Pulse Coupled Neural Network (PCNN) is introduced in this research field for its interesting properties in image processing  including segmentation  target recognition et al.  and a novel algorithm based on PCNN and Wavelet Transform for Multi-focus image fusion is proposed. First  the two original images are decomposed by wavelet transform. Then  based on the PCNN  a fusion rule in the Wavelet domain is given. This algorithm uses the wavelet coefficient in each frequency domain as the linking strength  so that its value can be chosen adaptively. Wavelet coefficients map to the range of image gray-scale. The output threshold function attenuates to minimum gray over time. Then all pixels of image get the ignition. So  the output of PCNN in each iteration time is ignition wavelet coefficients of threshold strength in different time. At this moment  the sequences of ignition of wavelet coefficients represent ignition timing of each neuron. The ignition timing of PCNN in each neuron is mapped to corresponding image gray-scale range  which is a picture of ignition timing mapping. Then it can judge the targets in the neuron are obvious features or not obvious. The fusion coefficients are decided by the compare-selection operator with the firing time gradient maps and the fusion image is reconstructed by wavelet inverse transform. Furthermore  by this algorithm  the threshold adjusting constant is estimated by appointed iteration number. Furthermore  In order to sufficient reflect order of the firing time  the threshold adjusting constant is estimated by appointed iteration number. So after the iteration achieved  each of the wavelet coefficient is activated. In order to verify the effectiveness of proposed rules  the experiments upon Multi-focus image are done. Moreover  comparative results of evaluating fusion quality are listed. The experimental results show that the method can effectively enhance the edge details and improve the spatial resolution of the image. 2011 SPIE.   System identification of tracking error and evaluation of tracking performance using BP neural network (EI CONFERENCE) 会议论文 International Symposium on Photoelectronic Detection and Imaging 2009: Advances in Infrared Imaging and Applications, June 17, 2009 - June 19, 2009, Beijing, China Zhang N.; Shen X.-H. 收藏  |  浏览/下载：14/0  |  提交时间：2013/03/25 A novel approach for evaluating the tracking performance of optoelectronic theodolite is proposed. First  an equivalent mathematic model of tracking error is established. Then  the equivalent sine signal is inputted to the equivalent model  and the outputs are sampled. The results of evaluating the tracking performance are obtained based on the statistical calculation of output produced by equivalent model. Equivalent model using the BP (Backprogration) neural network structure is identified. The training method of BP neural network adopts the LM (Levenberg-Marquardt) algorithm for the sake of speeding up training process. The BP neural network is trained and tested by using the training and testing samples gotten from the simulation model of optoelectronic theodolite tracking system under MATLAB/SIMULINK. The estimate errors of equivalent model including average error  maximum error and standard error are 2.5872e-0060  2.8 and 1.9. The results show that the equivalent model identified based on BP neural network meets the needs of evaluating the tracking performance of optoelectronic theodolite. The accurate evaluation of tracking performance is achieved. 2009 SPIE.   A MLP-PNN neural network for CCD image super-resolution in wavelet packet domain (EI CONFERENCE) 会议论文 2008 International Conference on Wireless Communications, Networking and Mobile Computing, WiCOM 2008, October 12, 2008 - October 14, 2008, Dalian, China Zhao X.; Fu D.; Zhai L. 收藏  |  浏览/下载：57/0  |  提交时间：2013/03/25 Image super-resolution methods process an input image sequence of a scene to obtain a still image with increased resolution. Classical approaches to this problem involve complex iterative minimization procedures  typically with high computational costs. In this paper is proposed a novel algorithm for super-resolution that enables a substantial decrease in computer load. First  decompose and reconstruct the image by wavelet packet. Before constructing the image  use neural network in place of other rebuilding method to reconstruct the coefficients in the wavelet packet domain. Second  probabilistic neural network architecture is used to perform a scattered-point interpolation of the image sequence data in the wavelet packet domain. The network kernel function is optimally determined for this problem by a MLP-PNN (Multi Layer Perceptron - Probabilistic Neural Network) trained on synthetic data. Network parameters dependent on the sequence noise level. This super-sampled image is spatially Altered to correct finite pixel size effects  to yield the final high-resolution estimate. This method can decrease the calculation cost and get perfect PSNR. Results are presented  showing the quality of the proposed method. 2008 IEEE.