Fixation based active visual segmentation and Restoration in Polar Space

Abstract

Image segmentation and restoration is a challenging task in the Computer Vision (CV). Image segmentation is the process of dividing an image into set of pixels and extracting the regions of interest. In recent times, there is an increasing interest in the use of segmentation to simplify and change the representation of an image into more meaningful and easy to analyze the model. This research work addresses three problems, fixation based segmentation, detecting repetitive pattern in the degraded image and restoring the degraded image. The main issues in the first problem are, there is no fixation point that lies inside a particular region of arbitrary shape and size. Only outer boundary segmentation is done. The exact number of regions are not identified accurately. newlineTo overcome the issues in fixation based segmentation, an interactive Segmentation by Weighted Aggregation (SWA) algorithm has been proposed. Segmenting a fixated region is equivalent to finding the optimal , closed contour around the fixation point. This closed contour should be a connected set of boundary edge pixels (or fragments) in the edge map. However, the edge map contains both types of edges, namely, boundary (or depth) and internal (or texture/intensity) edges. In order to trace the boundary edge fragments through the edge map to form the contour enclosing the fixation point. It is important to differentiate between the boundary edges from the non-boundary (e.g. texture and internal) edges. Incorporating fixation into segmentation has not only made the problem of general segmentation easier but also more robust. newlinevi newlineSecond problem is detecting a repetitive pattern in degraded image. To detect a repetitive pattern in an image. A suitable Pairwise Matching (PM) algorithm has been proposed. This is a well-defined issue to find an excellent solution to this problem. Detection of such repetitive structures can be used in many applications including image retrieval. The benefits of the proposed method are analyzed and evaluated with different parameter sets like precisions, recalls, throughput and F-measures. newlineFinally in the last problem, the resulting contours after segmentation and detection can be used to restore the weather degraded and noisy image with the help of Convolution Neural Network (CNN) model. An image restoration technique issued to reduce noise and recover resolution loss in the degraded image. To restore the noisy and degraded image an efficient fractional max pooling with ReLu (Rectified linear) in Deep Convolution Neural Network (DCNN) model have been designed and implemented. Deep Convolution Neural Network has a great deal of achievement in high-level vision problems such as object recognition, restoring weather degraded and noisy image. The experimental result is compared with different benchmarks like karapathy, MNIST, ILSVRC, VGG (Visual geometric group) and CIFAR 10 dataset. newlineIn the proposed Deep Convolution Neural Network CuDNN, a tremendous improvement over GPU computation has been done and it is optimized between 0.3GHz - 2.0GHz. The strength of the proposed approach lies in the simplicity and on the applicability of the mechanisms to other image applications. newline