| As a new photography equipment, the light field camera can obtain four-dimensional light information of the light field through one shot. The light field data can generate focal stacks image from which the depth map can be extracted. Currently, the light field camera has been widely used in computer vision, such as image segmentation, object recognition, saliency detection, three-dimensional reconstruction and so on. However, most of the existing methods lead to large defocusing errors, which are not robust in practical usage. In this thesis, we combine the focus detection algorithm with colour information to obtain high-accuracy depth information through analysing the feature of light field imaging and refocusing. The main work of this thesis is listed as follows:(1) We summarize the theories of light field and imaging system. Furthermore, we analyse the feature of light field data and technology of refocusing.(2) We implement an edge detection algorithm to calibrate the wrong coordinate of center of microlenses. The experimental results show that the right refocusing image can be extracted from light field data by using our calibration algorithm.(3) By analyzing the Pros and cons of most focus detection algorithms, we develop a windowed focusing detection function based on gradient mean square error to extract the depth information. Then we calibrate the wrong depth-information by guiding of colour information of all-focus image. Finally, we synthesize the depth images with MRF to improve accuracy. Our results demonstrate that the colour-guided depth map extraction from light field based on focusness detection achieves higher accuracy with less noise. |
