An Improved Canny Algorithm for Corneal Topography in Placido

Orthokeratology (OK lenses) is a specialized, high oxygen-permeable, rigid corneal reshaping contact lens that has been demonstrated to be effective in controlling and slowing down the progression of myopia in young people. The application of OK lenses during nocturnal hours enables the cornea to undergo a transient modification, thereby mitigating or eradicating refractive errors, thereby enhancing unassisted vision. To ensure the accuracy and reliability of the fitting parameters provided by the doctor during the diagnostic process, it is imperative to employ a corneal topography with distinct edges and a high signal-to-noise ratio. This is a pivotal step in the fitting of keratoconus lenses. The doctor’s assessment of the corneal topography is informed by their professional knowledge and experience, forming the foundation of their diagnosis. Prior to the acquisition of corneal topography, it is imperative to extract features from ocular photographs through the implementation of edge detection methodologies. Presently, the integration of the Canny edge detection algorithm and deep learning is employed to achieve corneal image ring recognition, a method that exhibits a commendable detection effect but is encumbered by suboptimal generalization capabilities and diagnostic efficiency. The traditional Canny algorithm has been the subject of much research due to its limitations in eliminating pepper noise and producing unclear extracted edges. An improved Canny algorithm has been proposed that overcomes these shortcomings. The proposed algorithm first uses contrast-limited adaptive histogram equalization (CLAHE) to enhance the image to be detected. Then, it creates a smooth ring mask. Finally, it uses the inverse Fourier transform to restore the image and extract the edge features. Through a series of comparison experiments, it was determined that the enhanced algorithm demonstrates superior performance in comparison to the conventional Canny edge detection method. The enhanced algorithm has been shown to detect more edge details, accentuate the contour, and enhance image quality by 10% to 15%. The Canny algorithm’s strengths, including its capacity for robust self-adaptation and effective interference removal, are preserved.