Image Enhancement Algorithm for Cold Cathode Weld Defect Based on Improved Retinex Theory

To address the issues of low contrast and poor visual quality in cold cathode weld defect images, an image enhancement algorithm based on an improved Retinex theory is proposed. First, bilateral filtering is used to remove stripe noise in the image. Next, the guided filter in the improved Retinex algorithm is used to estimate the illumination component, thus obtaining the reflection component. The illumination component is processed using contrast-limited adaptive histogram equalization (CLAHE), while the local details of the reflection component are enhanced using an improved Butterworth high-pass filter. The illumination and reflection components are then fused in a proportional manner. Finally, the image is enhanced using an improved homomorphic filter, resulting in the enhanced image. The algorithm's effectiveness is evaluated using subjective and objective assessment methods, with results indicating good image enhancement performance and more visible details.