.png){kind=icon}
.png){kind=icon}
Return
Weighted full binary tree-sliced binary pattern: An RGB-D image descriptor
.png){kind=icon}
OA
AI
DOI：10.1016/j.heliyon.2020.e03751.png)
.png){kind=icon}
.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
We present an Algorithm to understand Inter-pixel similarity, which shall be observed in images with the help of a data structure Full Binary Tree. The Full Binary Tree has certain properties like every node must have 2 children or none. Based on this property of Binary Tree, the method of Sliced Binary Pattern is proposed. The inter-pixel similarity may be observed by converting any pixel information of an image within a block of size 3 x 3 to its binarized form, as the pixel information, whose similarity with neighboring pixel cannot be exploited, when it is in decimal form. Thus, we convert all pixel information within a block of size 3 x 3 to its binarized form then we compare the binary pattern of a central pixel with its 8-nearest neighbors. If there is a binary pattern match between central pixel and its 8-nearest neighbors of a block, we assign weights to it, where the weights are determined by the position of match that exist between central pixel and 8-other neighboring pixels of an image. This process helps in determining the inter-pixel similarity of 8-nearest neighbors with respect to central pixel of a block. Every block of 3 x 3 pixels is processed with this strategy to obtain the similarity between patterns in an image. The erected Weighted Full Binary Tree-Sliced Binary Pattern analyzes an image in RGB-Dimensions based on patterns of Inter-Pixel Similarity by tracing the similarity path. The proposed RGB-D texture based inter-pixel similarity addresses the verification of facial similarity. Further, the proposed WFBT-SBP has yielded a good classification accuracy of 77.4%, 77.3%, 77.98%, and 77.94% over a relations of F-S, F-D, M-S, M-D of KinfaceW-I and 76.89%, 76.72%, 77.01%, 76.99% over a relations of F-S, F-D, M-S, and M-D of KinfaceW-II respectively.
Keywords:
Computer science
Sliced binary pattern
Representation learning
Feature extraction
Classification
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.