Research On The Image Processing Method And Its Application Based Onwavelet And Compressed Sensing

With the development of technology, image has played more and more important role in terms of our social. Image noise is an the virtue factor affecting image quality. On the other hand, with the image data increasing, it is necessary to compress the images before it is storage and transmitted. So the image denoising and compression has become a particularly important part in image processing.Around the image processing technology, the situation of wavelet theory and compressed sensing are analyzed in this paper, the traditional denoising algorithms and compression algorithms are summarized, and an improved multi-threshold denoising based on wavelet packe and hybrid sampling based on the compressed sensing reconstruction algorithm are proposed..For denoising process is difficult to save the details, proposed an improved multi-threshold denoising algorithm based on wavelet packet, while the maximum removal of noise to try to save the image detail. Before denoising, the first will be the image preprocessing, wavelet packet decomposition of the image layer, the low-frequency part of the reconstruction called approximate image, then make the image of the approximate edge detection, recording the image edge position.After three layers of noise image wavelet packet decomposition, calculating the energy spectrum coefficients.Denoising to different period of different energy strategy, the average energy is greater than or equal to one third of the frequency band selection Stein threshold processing.For other frequencies because of its containing noise, selects the fixed threshold for processing.The last completion of the denoising image edges and image fusion after denoising, get the final image.In order to improve the compression ratio while ensuring the restoration accuracy of important information, proposed reconstruction algorithm based on hybrid sampling compression sensing. The images is divided into important areas and non-critical areas, different regions with different reconstruction strategies. For the important region, the use of a high sampling rate and high quality of the reconstructed orthogonal matching pursuit(OMP) algorithm, while the gray scale of the non-critical areas is set to zero, thus equal to increases the sparsity of the import an region; For non-critical region using the lower sampling rate and faster reconstruction of segment matching pursuit(St OMP) algorithm. Finally, the image fusion, getting the high quality of the important region, faster reconstruction time and greater compression ratio.Finally, an improved image denoising based multi-threshold based on wavelet packet and mix samples compressed sensing reconstruction algorithm applied to intelligent traffic, through the images collected traffic makes the simulation experiment.