研究目的
To introduce a weighted sum-based color filter array interpolation method using Taylor series cubic approximation for accurate pixel value prediction and to remove zipper artifacts generated during the interpolation process.
研究成果
The proposed weighted sum-based demosaicking method using cubic Taylor approximation and a post-processing method based on clipping significantly improves the performance of demosaicking. The system outperforms existing demosaicking algorithms in terms of both objective and subjective performance, with effective removal of zipper artifacts.
研究不足
The utilization of a large local window may include color components from other objects or regions, potentially affecting the accuracy of interpolation.
1：Experimental Design and Method Selection:
The methodology involves using a high-order Taylor series approximation for accurate pixel value prediction during the reconstruction of quincunx grid-sampled green channels and performing a weighted average-based interpolation in a large local window. A color difference model in a small local window is used for additional green value generation.
2：Sample Selection and Data Sources:
The McMaster (McM) dataset was used for evaluating and comparing the performance of the algorithms.
3：List of Experimental Equipment and Materials:
Intel Core i7-8700K CPU @ 3.70GHz running MATLAB R2018a and an EIZO's ColorEdge CG277 monitor for subjective image quality evaluation.
4：70GHz running MATLAB R2018a and an EIZO's ColorEdge CG277 monitor for subjective image quality evaluation.
Experimental Procedures and Operational Workflow:
4. Experimental Procedures and Operational Workflow: The process includes G channel reconstruction using cubic Taylor approximation, G channel enhancement using a color difference model, R/B channel interpolation based on the enhanced G channel, and post-processing for zipper artifact removal.
5：Data Analysis Methods:
Performance was assessed using image quality metrics (IQMs) such as CPSNR, FSIMc, S-CIELAB ΔE*, ZER, and CPU time.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
