3-D Shape Recovery From Image Focus Using Self Organizing Map

Most applications in range segmentation, computer vision, consumer electronics and video segmentation basically require 3-D structure of the bodies from their 2-D images using a passive optical method known as shape from focus. Recovering a precise 3-D shape from its corresponding 2-D images is a vital passive optical scheme that utilizes the focus information of each pixel to estimate the depth map of the 3-D object. Recovering an accurate and precise 3-D shape from its 2-D images stacks using the shape from focus architecture is a fundamental objective of 3-D shape recovery. The focus measure computation plays a vital role in the accuracy and robustness of the shape from focus algorithm as it computes image focus volume from the acquired dataset of 2-D images. The proposed shape of focus architecture is efficient, easy to compute, performs reasonably to variant noise sources and robust to the conditions when the noises are not Gaussian. The proposed scheme is deeply investigated in variant noises under ample real and simulated image sequences. Experimental results demonstrate that the suggested 3-D shape recovery approach is both efficient and effective to support our claims.