Wood Image Super-resolution Reconstruction Based On SCN-MSE And Image Recognition

In wood industry, wood identification by visual analysis is relatively straightforward, with computer more accurate recognition. And because of imaging environment and restricted factors such as sampling devices. people get the image resolution is not high, unable to accurately extract timber information.Super-resolution technology based on the study of the core idea is using way of learning, to get the high and low resolution images, the correlation of module for the restored image as the basis, to complete the image of the supplement and optimization. Is characterized by accurate reconstruction of the image, and maintain the characteristics of image, the image noise robust. Based on the framework of study, based on the restored image details, improve image visual effect as the final goal, put forward to remove the center of the field are variance method, research on wood image super-resolution reconstruction.The main content of the thesis are as follows:1. Double focus on three interpolation algorithm, the interpolation algorithm based on adaptive direction, maximum a posteriori probability algorithm and the reconstruction of the sparse representation of the four classic super points algorithm theory.2. In the super-resolution reconstruction algorithm based on learning, on the basis of super-resolution reconstruction algorithm based on SCN-MSE is put forward. Established for the unit with the module training datablease, building blocks (N1-LBP operator is used to estimate coefficient, classifying module. Reuse the SCN-MSE operator, debris in LL finish module similarity comparison. Keep low frequency part of the low resolution imaged, the high frequency part of training in the datablease matching module, insert a low-resolution images, replace the high frequency part. Through the wavelet inverse transformation to complete the super-resolution image reconstruction.3. The SCN-MSE algorithm with double three interpolation algorithm, interpolation algorithm based on adaptive direction, maximum a posteriori probability algorithm and sparse representation through experiment contrast, and has carried on the subjective and objective evaluation to the result.4. Will the super-resolution reconstruction algorithm based on SCN-MSE recognition of wood, selection Mongolica wood and bark for object recognition, image super-resolution reconstruction based on SCN-MSE and image through traditional pretreatment. using SVM polynomial kernel function for identification, analysis and comparison the recognition result.