ベイヤー型カラーフィルターアレイのための確率的エッジ知覚モデルを用いた高周波成分の抽出

抄録

High-frequency waves, such as edges, and local distributions of intensity signal are important in the sense of material and textures. Color filter arrays, popularly adopted in digital cameras, capture only one of the R, G, or B values for each pixel. To capture full color data for each pixel, we have to interpolate the values from neighboring pixels. In this process, high-frequency waves, such as edges, may be lost in rough sampling and interpolation of color values. This paper proposes our new idea in interpolation of color values by probabilistically estimating 2D edge structure in high-frequent part to keep the original sense of material. Our basic idea is based on restoration of continuous digital line patterns, which corresponds to edge models. This paper also shows our experimental results and the performance comparing with the original full color image and the conventional method such as Hamilton method.