Edge Detection and Object Tracing in Real Image Sequences Using Gaussian Approximation with Graph-Cut Model
M. Sivakkumar, Santhosh Satheesh · 2013
In image segmentation process, edge enhancement is the important process where edge information can be utilized to acquire partitions analogous to actual objects, or significant parts of the objects in the image. Edge enhancement is normally done through filtering process. The filter can be employed as a pre-processing implementation in image segmentation. The previous work described only the image segmentation process using vector filtering approach for both layered and featured image and segment the image as vector field. The drawback of the previous work is that it did not involve in the edge enhancement process and had no technique to trace the moving object patterns in the segmented image to identify the motion and direction of object in the real image sequences. To enhance the process, in this work, we present a new adaptive nonlinear filter intended at smoothing the edges of the given real image sequences. Several special features are introduced to the filter, including Gaussian approximation with graph cut model to Evaluate the two or more image sequences with higher inter frame motion. Non-linear smoothening Filters are first applied on the pixels of the image to enhance the edge detection process. After non- linear filtering followed by image segmentation, the Tracing of growing/moving patterns of object is identified using Gaussian approximation with graph-cut model. An experimental evaluation is conducted with real image sequences samples to evaluate the effectiveness of the proposed edge detection and object tracking using Gaussian approximation with graph-cut model in terms of edge localization, object tracing speed, computation time for object tracing.