Visual Motion Detecting and Deblurring Based on Mathematical Morphology and Ensemble Learning

The problem of blurring caused by object motion in a gray level image is analyzed, and an algorithm combining image segmentation and blind deconvolution based on statistical features of object and background is introduced to estimate visual motion and restore the image. Suspected regions with slowly changing intensity of pixels are segmented on the base of gradient and curvature of the image. Simple connected regions are selected by the use of mathematical morphological algorithm, and convolution kernels of regions larger than a given threshold are inferred through ensemble learning. Motion patterns of objects can be determined and the blurred region can be restored. Experimental results show the effectiveness of the algorithm for visual motion estimation and deblurring in a gray level image.