Pulse coupled neural network based image classification

A Hardware-Implementation-Friendly Pulse-Coupled Neural Network Algorithm for Analog Image-Feature-Generation Circuits

“…Pulse-coupled neural networks (PCNNs) are biologically inspired algorithms initially introduced for modeling a cat's visual cortex and developed for high-performance biomimetic image processing. [1][2][3][4][5][6][7][8][9] Research on PCNNs over the past decade has shown that PCNNs are highly applicable in the field of image recognition, particularly in image feature generation. [3][4][5][6][7][8][9] When applied to image feature generation, PCNNs have significant merits such as robustness against noise and independence of geometric variations in input patterns.…”

“…[1][2][3][4][5][6][7][8][9] Research on PCNNs over the past decade has shown that PCNNs are highly applicable in the field of image recognition, particularly in image feature generation. [3][4][5][6][7][8][9] When applied to image feature generation, PCNNs have significant merits such as robustness against noise and independence of geometric variations in input patterns. 4,8,9) However, conventional PCNNs [3][4][5] are software-oriented algorithms containing a large number of convolution operators and floating-point multipliers, which are too complicated to implement as compact very-larege-scale integration (VLSI) hardware.…”

“…[3][4][5][6][7][8][9] When applied to image feature generation, PCNNs have significant merits such as robustness against noise and independence of geometric variations in input patterns. 4,8,9) However, conventional PCNNs [3][4][5] are software-oriented algorithms containing a large number of convolution operators and floating-point multipliers, which are too complicated to implement as compact very-larege-scale integration (VLSI) hardware. [10][11][12] To employ PCNNs in image-feature-generation VLSIs, a hardware-implementation-friendly PCNN algorithm is proposed here.…”

A new feature extractor invariant to intensity, rotation, and scaling of color images

“…Therefore, implementing the third-order neural network technique in realworld applications is not feasible. A pulse-coupled neural network [9,14,23] can be used to cope with image segmentation with invariant properties. Zernike moment invariants introduced by Khotanzad [15] are used in most applications requiring invariant properties [19].…”