XCR (eXponential-Cos Retinex)
A fast, structure-preserving Retinex method based on a complex exponential kernel, released as a preprint on Jxiv.
- Year
- 2025
- Role
- Sole author
- Category
- Research
- Status
- Completed
- Links
XCR is a fast, structure-preserving method for low-light image enhancement (Retinex) that estimates illumination with a kernel based on the complex exponential. The paper, “A High-Speed and Structure-Preserving Retinex Method Based on a Complex Exponential Kernel: XCR” (in Japanese), was released on Jxiv in November 2025.
I refined it through mentoring in LEADING EDGE Shikoku (part of METI’s AKATSUKI Project) and presented it together with FSSR at NUS (National University of Singapore) during the overseas programme in Singapore in November 2025.
A later paper showed that the XCR kernel equals the sigmoid derivative sampled on the imaginary axis, which makes it unstable at the poles of the logistic function. Its successor, SD-Retinex, fixes this.