Fusion Of Light Field Focus Stack And Full Focus Image Of The Significance Check Algorithm

Salient object detection has been a hot research topic in the field of computer vision in recent years. Accurate and reliable saliency detection is helpful to the subsequent fast implementation of object tracking and recognition. The existing saliency detection methods have obtained significant achievements. They usuallly integrates the low-level features (pixels or super-pixels) and high-level descriptors (regions or objects) together to realize the salient object detection. However, these methods are only suitable for the conventional cameras that are based on traditional integral imaging.Lytro can obtain light field data by introducing the micro lens array in front of the image sensor, which breaks through the limitation of the traditional camera, and further provides new ideas and methods for image processing and computer vision. And recently, the applications of light field camera have attracted a wide attention of many researchers around the world.In this paper, we mainly explored the theories of light field camera, the extraction and processing of light filed data, and the saliency detection of light field image. The main contents include:Firstly, by analyzing the imaging principals and algorithms of light field camera, we proposed a set of methods for light field data extraction and decoding.On this basis, we further focusd on the transform algorithm from hex to rectangle for Lytro image decoding, which enhances the accuracy of light field data extraction.Then, according to the characteristic of light field camera, we put forward an approach to fuse the focus stack images obtained from the Lytro camera. And this new fused image is defined as an all-in-focus image.Lastly, we developed a new saliency detection algorithm based on the focal stack and the all-in-focus image, which can robustly handle challenging scenarios.