Texture Gradient Suppression Strategy And Structure Pixels Detection Method For Texture Filtering

Texture filtering is to filter out the texture details while preserving the important structural line information in the image,but it is very difficult to achieve these two contradictory indicators.Texture filtering is a fundamental and important research topic in computer graphics and digital image processing,and has important practical applications in such fields as image editing,tone mapping,high dynamic range compression.In this paper,texture filtering algorithm is studied in three aspects:(1)Aiming at the problem that strong gradient texture can't be suppressed,this paper proposes a texture filtering algorithm combining texture gradient suppression and L0 gradient minimization.Firstly,we improve the directional semicircular gradient operator,which has the capability to distinguish texture and structure pixels.The local contrast stretching and scale adaptive strategies are adopted to improve the recognition ability of weak gradient structure pixels.Subsequently,the original image gradient is suppressed by the directional semicircular gradient amplitude,and the image reconstruction is performed with the suppressed image gradient to obtain a texture suppression image whose texture pixel gradient is smaller than the structural pixel gradient.Finally,we use the L0 gradient minimization algorithm with gradient lifting effect to filter the reconstructed image to remove the texture while preserving the structure because the texture gradient suppression operation causes a certain attenuation of the gradient of the structural pixels and thus results in the loss of the structure.The experimental results show that the proposed algorithm can maintain the main structure of the image while suppressing the strong gradient texture,and has good universality and robustness.(2)Aiming at the problem that the fine structure can't be maintained and the virtual edge is difficult to generate,this paper proposes a texture filtering algorithm that conforms to visual perception guided by structure detection.The method first calculates the directional semicircular gradient and suppresses its amplitude non-maximum value to obtain preliminary structural detection results.Subsequently,the false edge suppression strategy is used to remove the burrs and isolated points in the preliminary detection results,and the structural correction method is used to correct the structural pixels that deviate from the original position,and the lost complex structure is recovered by the structural complementation operation.After the above three steps of processing strategy,we can get a detailed structure inspection map.Finally,the directional semicircular gradient is given to the structure diagram to obtain the ideal gradient field,and the gradient image is used to reconstruct the filtered image.The experimental results show that the proposed algorithm can maintain weak gradient structures and complex structures while suppressing strong gradient textures,and generate virtual edges that conform to human visual perception.(3)Aiming at the difficulty of manual design features,this paper proposes an adaptive bilateral texture filtering algorithm based on depth structure prediction.Firstly,a deep structure model based on dense network and wing loss function is proposed,which can predict the structural probability value of each pixel well and reach 94.26%accuracy in the structured/unstructured dataset manually collected and data augmented.Then,the optimal filter scale of each pixel is estimated by using the structural probability value,and the guided image whose texture is mostly suppressed is obtained by the scale-adaptive Gaussian filtering.Finally,combined with the bilateral texture filtering method based on the guide graph,the filtering result of the texture suppression structure is obtained.The experimental results show that the depth structure model proposed in this paper can effectively predict the structure probability,and combined with the scale-adaptive bilateral texture filtering method,it can suppress the texture and maintain the main structure of the image.