Self-adaption image weighted mean filter design based on FPGA

Abstract

In order to solve the problem of noise preprocessing in the process of image acquisition by complementary metal oxide semiconductor(CMOS) image sensor, a configurable adaptive weighted mean filter module in FPGA is proposed. This module determines the direction of the texture by detecting the variance of different directions in the filter window, thus automatically generating the corresponding weighting coefficients. The weighted mean filter can process image with a width of no more than 4 094 pixels, so as to achieve the goal of de-noising and preserving edges. Finally, the experiment shows that when the image width is 2 048, the delay is 42.04 μs, and it satisfies the real-time requirement. The image noise is obviously reduced. The texture edge burr is disappearing, and the image quality can be improved better after filtering.

Keywords

• FPGA
• CMOS image sensor
• mean filtering
• fixed noise
• logic design