A new FPGA accelerator based on circular buffer unit per orientation for a fast and optimised GLCM and Texture feature computation

Mohamed Amin Ben Atitallah (1) , Rostom Kachouri (2) , Hassene Mnif (1)

1 LETI - Laboratoire d'électronique et des technologies de l'Information [Sfax]
2 LIGM - Laboratoire d'Informatique Gaspard-Monge

This paper presents an FPGA accelerator based on circular buffer unit per orientation for a fast and optimized Gray Level Co-occurrence Matrix (GLCM) and four Texture features computation. The Four texture features namely, contrast, energy, dissimilarity and correlation are computed using Xilinx FPGA. However, the computation of GLCM and four textures features are very complex and consume a lot of execution time. In this paper, an FPGA accelerator for fast computation of GLCM and four texture features are designed and implemented. This architecture was implemented on a Xilinx Zc-702 using Vivado HLS. The obtained results are then compared against other related works. The synthesis results on FPGA prove a significant gain (about 17%) in execution time compared to the previous work.

Mots clés: Image analysis applications; Parallel calculation; Haralick's texture feature; Hardware/Software Implementation; Circular buffer unit; FPGA; Vivado_HLS; Optimization; Execution time

Domaines: Architectures Matérielles [cs.AR]; Systèmes embarqués

Mohamed Amin Ben Atitallah, Rostom Kachouri, Hassene Mnif. A new FPGA accelerator based on circular buffer unit per orientation for a fast and optimised GLCM and Texture feature computation. IEEE international conference on Design & Test of integrated micro & nano-Systems (DTS), Apr 2019, Gammarth, Tunisia. ⟨hal-02172154⟩