A highly aggressive disease of the female reproductive system, ovarian cancer, is frequently discovered at an advanced stage because of vague symptoms including pelvic pain and abdominal edema. Although diagnostic imaging has advanced, there are still difficulties in improving histopathological pictures for precise classification, especially in the areas of noise reduction and structure preservation, which are crucial for deep learning models. The usefulness of several denoising filters, such as Gaussian, Median, bilateral, Wiener, and non-local means, in improving histopathological pictures for the diagnosis of ovarian cancer is examined in this work. For deep learning-based categorization, the filters were chosen for their capacity to minimize noise while maintaining important image structures. According to quantitative study, the bilateral filter fared better than all others, obtaining low mean square error (0.00160) and root mean square error (0.0398) while having the highest peak signal-to-noise ratio (28.06), structural similarity index method (0.60) and signal-to-noise ratio (25.89). In contrast, Wiener filter performed the worst with greater MSE (0.01016) and RMSE (0.0956), as well as PSNR (20.72), SSIM (0.43) and the SNR (18.55). These findings demonstrate the crucial filter selection is to a precisely diagnosis. By increasing the accuracy of deep learning models, this research work may improve the diagnosis of ovarian cancer. Further studies may investigate integrating these diagnostics techniques into clinical process for real-time diagnostic support. By concentrating on optimal denoising approaches designed for ovarian cancer imaging, this study closes the gap and offers important insights into enhancing ovarian cancer classification by optimal picture pre-processing. This study fills the gap by focusing on optimal denoising techniques tailored for ovarian cancer imaging, providing valuable insights into improving ovarian cancer classification through optimized image pre-processing.