Upgrading to a soft multifunctional image processor for implementation on a field programmable gate array with additional biasing and logical capabilities

Abstract

In this paper, there are details of the continuation of the development of the previously designed multifunctional morphological system and a description of how the system has been upgraded by including soft morphological operators as their ability to compensate for noise by varying the k-value of the system and how this is of great importance in image processing. This paper will cover the details of the upgrade process to include a greater number of logical functions than its predecessor, how a biasing system was included to brighten or dim a certain quadrant of the image or the entire image, how images are combined in either a product, mathematical or logical block and finally how the control mechanism is interfaced so that there are no signal clashes and that complicated strings of possibly many merged or biased images can be processed and outputted. The testing process will be discussed where each layer is tested for k-value and extra operations which are then compared to MATLAB test images from the Image Processing Toolbox or hand written m-files as a basis for comparison. The paper will also include how certain areas have been made more efficient to make the system faster and to reduce space on the FPGA by reducing the number of convolution kernels compared to the latter.

Image processing virtual circuitry for physics based applications on field programmable gate array technology

“…The latter change provides an additional advantage that enables the user to read the value quickly rather than converting a number to decimal from the bit extractors, hence making any modifications easier. This change was also employed in [17]. Now that the control circuitry for the selection is in place, the circuitry follows the same principle as in [16] but now the selection process is governed by the upper and lower limits shown in Table 3 and visualised by the rounded boxed area in Figure 5.…”

“…This presented two possible options, to continue to track the selected region after selection to ignore a vast amount of the movement, or to create a point of grayscale '255' at the centre of the image to show the movement. The first option is preferred as the latter option involves splitting the image into four segments to calculate the movement and insert the points simultaneously and then recombine them, previous work [13,15,17] has shown that the computational complexity and resource requirements are extremely high for an FPGA based system while the circuitry also runs the risk of incurring a propagation delay and other associated digital electronics problems.…”

“…As certain pixel values in Table 3 are shared by more than one colour, the pixel extraction technique has been expanded to include spatial measurements in addition to the value measurements. The pale area at the bottom of Figure 5 are x and y coordinate counters from [17] so that the system knows its current location in the image being considered. This system has an additional feature shown by the boxed in area, thankfully the two similar areas are in different parts of the image, one near the top and the other at the bottom.…”