Towards Quantitative Measures of Image Morphing Quality

Image morphing has been successfully used to synthesize realistic transitional effects in the computer graphics field for more than 15 years. It has a wide range of applications from the entertainment industry to human machine interfaces. However, little work has been done to quantify the quality of the morphed transition. Consequently, there are no quantitative measures that allow one to compare among different morphing approaches. Our work aims to quantitatively address this particular question — "Is this a good morph?". This paper presents an overview of our ideas and describes an error measure based on epipolar geometry for morphing between images of different viewpoints.