PROJECT TITLE :

Graph-Regularized Locality-Constrained Joint Dictionary and Residual Learning for Face Sketch Synthesis

ABSTRACT:

For digital entertainment and police enforcement, face sketch synthesis is a critical issue It's a great way to bridge the gap between face photographs and doodles in terms of texture. We refer to these as "unusual qualities" since they are difficult for current face sketch synthesis algorithms to understand from the relationship between pictures and sketches. Using residual learning, we provide a novel face sketch generation approach. With our method, we want to forecast the residual picture from a photo by learning the mapping relationship between a photo and its residual, which is the difference between a photo and its sketch, as opposed to traditional approaches, which aim to recreate a sketch image directly. In order to optimise the residual mapping, this method will make it easier than optimising the original mapping and obtaining unusual characteristics. Additionally, we offer a joint dictionary learning approach that preserves the local geometry structure of a data space. Face sketch synthesis is transformed from an image space to an entirely new and compact one, which is spanned by learned dictionary atoms, thereby ensuring the manifold assumption. Proposed methods show excellent performance in face sketch synthesis on three public datasets and numerous real-world pictures, according to the results. These findings are based on comparing the proposed method to a number of current methods, including some recently proposed deep learning-based approaches.