Disparity estimation is an essential step for 3D depth- based processing. Generally, it is categorized as either local or global method. Local methods are structurally simple and very efficient in real-time processing but they tend to produce relatively poor quality on homogeneous and noisy environments. We present an accurate and efficient one-pass local disparity algorithm that is robust to image noises and a variety of texture regions. Moreover, the proposed method is extended to the video domain by using motion as well as imposing temporal consistency. Large panoramic views have advantages such as wide field of view and high resolution. However, the large size poses a challenging problem for many computer vision tasks. We present an effective multi-resolution depth and fusion algorithm with adaptive disparity search range that achieves better accuracy and reduces complexity at the same time. Multi-array cameras have greater potential for many vision applications than stereo cameras. However, there are very few research results due to lack of data. We create multi-array dataset with associated ground-truth disparity so that other researchers can use for comparison. We present a new cascade regularization approach, which regularizes diagonal variations better and, in turn, yields better disparity enhancement as well as image restoration. The proposed disparity enhancement algorithm achieves performance gain of up to 65% compared to the initial disparities