Passive Millimeter Wave Image Adaptive Denoising And Sparse Reconstruction Algorithm Study

Passive millimeter-wave (PMMW) imaging is a method of forming images throughthe passive detection of differences of the millimeter-wave radiation energy betweenscene and the targets. In our PMMW imaging system, due to the limitations of systemantenna aperture and noise interference, the quality and resolution are not enough forour practical application. And yet the general approaches have ignored the features ofPMMW images themselves. So it is very necessary that to work on the specific methodfor the practical PMMW image processing.Concentrating on the PMMW image denoiseing and reconstruct processing, thisthesis focus on the PMMW images adaptive sparse decomposition, and adaptivedenoising methods based on the PMMW images features. The main contents include:1. During the pro-processing of PMMW images, comparing that median filteringhas blurred the details when removing noise, we proposed a impulse noise detectivealgorithm based on median filtering, and give the principle of choose detectivethreshnold. This algorithm has kept the denoising ability of median filtering way, butavoided unnecessary images details blurring of traditional way.2. Aim on the misjudging the metal point target as impulse noise, we analyze theimage differences’ differences between target and noise, and proposed a new adaptivenoise detect algorithm base on normalized image differences. It has further improve theability of noise removing, and help target detection and recognition in followingprocedure.3. We analyzed the sparsity of PMMW images based on over-complete dictionary,and studied the local sparse deposition model and algorithm based on patches. And thenwe used K-SVD algorithm to learn over-complete dictionary for PMMW images,showing a good sparse representation capability.4. According to the local sparsity of PMMW images, we analysed a PMMWreconstruct model based on patches. We introduce the sparsity on over-completedictionary to the constrict term of iteration threshold algorithm. By using thenon-negative finite value information of PMMW images, it shows a good effect when noise existed, and perform a good reconstruct of PMMW images.At last, we applied our denoising and reconstruct algorithms to both simulatedPMMW and in practice image processing of PMMW imaging system, and showed asignificant results.