Abstract:To address the shortcomings of the autofocus?window construction technology used in conventional focusing methods for laser processing, such as limited real?time performance, vulnerability to background noise, and considerable computational complexity in microtexture applications, this study proposes a saliency?window construction method. This method utilizes image saliency to isolate the region of interest for focus evaluation by integrating gradients and color cues. The main idea is to apply a visual saliency?detection framework when generating an autofocus window to suppress irrelevant background information and enhance the specificity of sharpness evaluation. A window?extraction algorithm combining Kirsch gradient analysis, LAB color?contrast calculation, local normalization, and texture suppression is introduced, thus enabling microtexture regions to be reliably depicted under brightness variations and complex modes. The method was experimentally verified on a laser microprocessing device. Compared with the conventional central?window method, the proposed method exhibits higher levels of sensitivity, steepness, and sharpness ratio by approximately 83%, 75%, and 54%, respectively, thereby enabling the generation of single?peak and fast?response focus curves and supporting stable microstructure manufacturing. Owing to its ability to balance between high precision and real?time performance, this saliency window provides an effective solution for the laser microprocessing platform.