FPGA Implementation of Image Block Generation and Color Space Conversion for the Gaussian Mixture Model

Abstract

Due to high rise in demand for video processing on hardware FPGA implementation of various image processing algorithm has become a necessity. Lot of research work is being done in the field of image processing and video processing on FPGA. In this paper we specifically concentrate on working Background identification, moving object detection and tracking. Keeping in mind the previous work dons in this area we have selected Gaussian mixture model for the implementation. The main work carried out by Gaussian mixture model is to is a common feature in many video processing systems. The paper speaks about two block of implementation mainly the image block generation and the RGB to YCbCr conversion. It has been proved from the design that it is efficient for high definition (HD) video sequences with frame size 1920 × 1080. Both the bocks are targeted on commercial field-programmable gate-array (FPGA) devices. When implemented on Atrix7 100t it uses 3% of FPGA logic resources. The second block takes like 2% of the FPGA resource. The total time required to process single frame is less than 2ns.