An improved demosaicing method based on Hibbard algorithm

Abstract

As computer vision become more widespread and deeper, the requirements for accuracy and clarity of an image gradually increase, and more accurate image reconstruction algorithms urgently need to be developed and researched. Existing image reconstruction algorithms mainly use the conversion of images into RGB or use the difference to predict certain regions to build mathematical models and then revert to images. The existing bilinear interpolation algorithm may reduce the resolution and damage the high frequency part, thus blurring the edges and making it difficult to achieve the expected results. The colorimetric constant method is greatly affected by noise, and the boundary problem is still unsolved. The gradient edge interpolation method only considers the horizontal and vertical directions, which is a single idea and does not meet the needs of oblique edges, and there will be problems such as color overflow. And based on some features near the oblique boundary, this paper focus on the case of large or small edge slope, and select the horizontal difference and vertical difference for algorithm improvement respectively by comparing the color difference along all four directions. Experimental results show that the PSNR value increases slightlyby 0.0013. It effectively optimizes the image and lays the foundation for feature perception in the image.