Image Super-resolution Reconstruction Based On Sparse Representation And Guided Image Filtering

People understand and record the objective world through images.With the continuous development of various advanced video recording equipment,how to make the low resolution images or videos clearer through software technology without changing the image acquisition equipment has become a research hotspot.The technology of the image super-resolution reconstruction has been used in sensing imaging,medical imaging,video surveillance and other fields.In this paper,the image super-resolution reconstruction method based on sparse representation is studied,the high-frequency detail of the reconstructed image is insufficient and there is an artifact at the edge of the image.In view of this deficiency,this paper makes improvements from two aspects:First,an image super-resolution reconstruction method based on residual dictionary is proposed in this paper.The low-resolution images in training set are reconstructed based on sparse representation,and corresponding high-resolution images are obtained,whose difference with the original high resolution image in the training set is named the residual images.By training the residual dictionary using the residual image features,and the high-frequency information lost in the acquisition process is restored.Second,this paper presents an image super-resolution reconstruction method based on sparse representation and guided image filtering,decomposing input image into the edge layer,the foreground color and the background color,among them the reconstruction of the edge layer is achieved through learning dictionary pair and two other images by bicubic interpolation,the final high resolution image is obtained by image synthesis.The images reconstructed by this method are better than the traditional image reconstruction via sparse representation in objective evaluation indicators after experimental verification.From the visual effect point of view,the high-frequency edges of the image reconstructed by this method are more detailed and the color information is closer to the original high-resolution image.