Optimal Pyramid Column Feature with Contrast Enhanced Model for Face Sketch Synthesis

Narasimhula Balayesu and Hemantha Kumar Kalluri

In this paper, we present a new method for synthesizing a face sketch from a photo using deep neural networks. The face sketch has been synthesized by the framework through replicating the artists form sketch in a cascading way. Before transforming the digital image to face sketch, gamma correction is applied to enhance the contrast of the image. Next, the content image is produced which make face shape outline and key facial features. To improve the sketch details, shadings and textures are inserted. To generate a content image, fully convolutional neural network (FCNN) is employed first and then a style transfer method is applied to set up shadings and textures depending on the new projected pyramid column feature with gamma correction (PCF-G) method. The style transfer strategy preserves additional sketch details that depend on the pyramid column feature when comparing with general style transfer strategy and conventional patch-based methods. Qualitative and quantitative examinations recommend that this structure is even better when compared with the standard techniques on the applied various sample images. The presented PCF-G method exhibits superior results with a maximum Structural similarity index (SSIM)value of 0.504 on the applied test images.

Volume 11 | Issue 5

Pages: 335-344