Visual quality assessment of the synthesized sketch

The existing image and video quality assessment algorithms are devoted to the degradation of quality caused by some artefacts or noise. We propose a novel image quality assessment for the synthesized sketch. When the sketch is synthesized from a photo, the quality of the resulted sketch is evaluated from two aspects. First, the synthesis of sketch may be affected by artefacts and therefore its visual quality should be measured with no-reference quality evaluation algorithm; secondly, the synthesized sketch should preserve discrimination information as the corresponding photo, and their similarity is also examined in the proposed method, with the photo as reference image. Because sketch synthesis aims at enhancing the performance of sketch-photo recognition and the purpose of sketch quality assessment is to determine whether the quality of the resulted sketch is sufficient for sketch-photo recognition, the proposed sketch quality assessment is associated with the recognition rate of sketch-photo recognition, that is to say, the rationality of the proposed method is judged according to the performance of sketch-photo recognition. A set of experiments are performed to prove that our method is more valid than the traditional methods.