Image Details Enhancement Method via Spectrum Total Variation

Abstract

In order to enhance the texture details of images and preserve the complete structure of images, we propose a novel image detail enhancement method based on spectral total variation, which combines the total variation method and spectral analysis method. Firstly, the total variation flow, as a non-ideal high-pass filter, decomposes image and obtains the spectral characteristics of different time scales in the transform domain. The transform domain can be regarded as a frequency domain. Then, we construct a special filter in the transform domain, and the spectrum features are filtered and enhanced by the filter. Finally, according to the proposed reconstruction method, the filtering results in the transform domain are inversely converted to the spatial domain, that is, the final enhanced image is obtained. The experimental results show that the proposed spectral total variation method can effectively use the spectral information of the image to enhance the local details, so that the enhanced image can obtain more prominent texture information, while maintaining the integrity of the image structure.