Fusion-Based Image Enhancement and Its Applications in Mobile Devices

In this chapter, we introduce a general framework for fusion-based image enhancement to produce photorealistic or non-photorealistic enhancement effects from natural images. For example, based on the contrast adjustment technique, we can produce the naturalness-preserving low-light enhancement effect for dark or back-light images. Also, based on the pencil drawing stylization and the semantic segmentation technique, we can convert a natural image into the mixed reality-virtuality effect. In these applications, we adopt a simple-but-effective method to construct a pixel-wise fusion map, aiming to guarantee the fusion-based enhancement effects. Furthermore, we have realized the techniques in mobile devices. For better user experiences, we use Fast Fourier Transform (FFT) to accelerate some computationally expensive steps. In addition, we introduce GrabCut to allow users to extract any interested regions for generating the mixed reality-virtuality effect. Demo APPs for Android phones are available for downloading.