Worked on optimizing the Fluid Select kernel in the opencv/opencv repository, focusing on enhancing performance and accuracy for image processing tasks involving multi-channel (2/3/4) and 16-bit data. Leveraged C++ and OpenCV’s Universal Intrinsics to migrate the SIMD path from legacy implementations, introducing branchless execution and improved mask handling for mixed bitwidths. Updated and expanded accuracy and performance tests to validate the new kernel path, ensuring stability and quantifiable throughput gains. The work emphasized efficiency and compatibility across common channel and bitwidth configurations, laying the foundation for future SIMD enhancements without addressing critical bug fixes during this period.