Low Complexity Background Subtraction for Wireless Vision Sensor Node

Energy-Efficient SRAM FPGA-Based Wireless Vision Sensor Node: SENTIOF-CAM

“…Therefore, we have considered these parameters in relation to selecting a suitable scaling technique for both upscaling and downscaling. Based on experimentations [35], it was concluded that averaging for downscaling and nearest neighbour for upscaling, with a scaling factor of 8, are suitable for the machine vision applications. The scaling factor of 8 means scaling in both width and height.…”

“…For upscaling, the nearest neighbour is used. This method requires the sampling of the nearest pixels to the original image [35] and the implementation is straightforward.…”

“…Some of these techniques such as the Gaussian method, offer good quality but have higher computational complexity and require an approximately 22 times longer execution time as compared to that for the bilinear [33]. For the proposed background model, different image scaling techniques, including nearest neighbour, averaging, bilinear, and bicubic with different scaling factors were first investigated for a real application [35]. In relation to image scaling techniques, the output image quality, complexity and memory requirement depends on two factors including scaling technique and scaling factor.…”

“…The contents that are lost, require the generating of the image after each wake up cycle, which is costly in terms of computation and energy consumption. It is important to mention that the background remains constant for a long time in many machine vision applications [5] [35]. To handle this situation, the background scaled down image is stored in the non-volatile FLASH for later use.…”

Low complexity FPGA based background subtraction technique for thermal imagery

“…There is no model addressing all these issues. Therefore a model should be investigated with careful considerations because of aforementioned constraints [10] [11].…”