Modified Local Binary Pattern Scheme using Row, Column and Diagonally Aligned Pixel’s Intensity Pattern
Nitin Arora1, Alaknanda Ashok2, Shamik Tiwari3
1Nitin Arora, Department of Computer Science & Engineering, Uttarakhand Technical University, Dehradun (Uttarakhand), India.
2Alaknanda Ashok, Department of Electrical Engineering, G. B. Pant University of Agriculture and Technology, Pant Nagar (Uttarakhand), India.
3Shamik Tiwari, Department of Cloud Computing & Virtualization, University of Petroleum & Energy Studies, Dehradun (Uttarakhand), India.
Manuscript received on 07 March 2019 | Revised Manuscript received on 20 March 2019 | Manuscript published on 30 March 2019 | PP: 771-779 | Volume-8 Issue-5, March 2019 | Retrieval Number: D2797028419/19©BEIESP
Open Access | Ethics and Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: This paper, suggested a novel method for texture feature descriptor using the row, column and diagonally aligned pixel intensity difference for efficient image retrieval based on image contents (CBIR). In Local Binary Pattern (LBP), a matrix of size 3×3 of an image is used, and then the central pixel of the window is subtracted from all its eight neighbours one by one pixel. LBP uses, 0 or 1 based on if the variance between central and neighbour pixel value is negative or positive respectively, to generate a binary pattern. Decimal value to this 0 and 1 binary string represents the binary value of the corresponding image pixel. By doing this, LBP ignores the effect of the row, column and diagonally aligned pixels of an individual pixel for its encoding to binary values and similarly for its texture explanation. This suggested texture descriptor is centered on the clue that row, column and diagonally aligned pixels of an individual pixel hold momentous quantity of data and this data that can be used for operational and proficient texture demonstration for CBIR. This method do not dependent only on the sign of the intensity values between central pixel and its neighbours as in pre-existing LBP methods, rather suggested technique measured the sign of variance between and its row, column and diagonally aligned pixels. Using this concept, in this paper we suggested a row, column and diagonally aligned pixels Intensity Pattern (RCDAPIP) based texture descriptor. This method considers the comparative intensity difference between a particular pixel and the center pixel by considering its row, column and diagonally aligned pixels and generate a sign (SRCDAP) and a magnitude pattern (MRCDA). Finally, both the patterns SRCDAP and MRCDAP are merged into single pattern (RCDAP) to produce a more proficient feature descriptor. The suggested technique has been tested on WANG database of one thousand images. The Euclidean and Manhattan distance are used for similarity measure. The precision % value and the recall % value is calculated on suggested technique are equated with existing local binary pattern (LBP). The suggested system indicated a momentous enhancement over pre-existing LBP technique.
Keyword: Extraction, Local Binary Pattern, Pixel Intensity, Texture Feature, Texture Content based Image Retrieval.
Scope of the Article: Pattern Recognition