Image Processing Strategies Based on a Visual Saliency Model for Object Recognition Under Simulated Prosthetic Vision

<jats:title>Abstract</jats:title><jats:p>Retinal prostheses have the potential to restore partial vision. Object recognition in scenes of daily life is one of the essential tasks for implant wearers. Still limited by the low‐resolution visual percepts provided by retinal prostheses, it is important to investigate and apply image processing methods to convey more useful visual information to the wearers. We proposed two image processing strategies based on Itti's visual saliency map, region of interest (<jats:styled-content style="fixed-case">ROI</jats:styled-content>) extraction, and image segmentation. Itti's saliency model generated a saliency map from the original image, in which salient regions were grouped into<jats:styled-content style="fixed-case">ROI</jats:styled-content>by the fuzzy<jats:italic>c</jats:italic>‐means clustering. Then<jats:styled-content style="fixed-case">G</jats:styled-content>rabcut generated a proto‐object from the<jats:styled-content style="fixed-case">ROI</jats:styled-content>labeled image which was recombined with background and enhanced in two ways—8‐4 separated pixelization (8‐4<jats:styled-content style="fixed-case">SP</jats:styled-content>) and background edge extraction (<jats:styled-content style="fixed-case">BEE</jats:styled-content>). Results showed that both 8‐4<jats:styled-content style="fixed-case">SP</jats:styled-content>and<jats:styled-content style="fixed-case">BEE</jats:styled-content>had significantly higher recognition accuracy in comparison with direct pixelization (<jats:styled-content style="fixed-case">DP</jats:styled-content>). Each saliency‐based image processing strategy was subject to the performance of image segmentation. Under good and perfect segmentation conditions,<jats:styled-content style="fixed-case">BEE</jats:styled-content>and 8‐4<jats:styled-content style="fixed-case">SP</jats:styled-content>obtained noticeably higher recognition accuracy than<jats:styled-content style="fixed-case">DP</jats:styled-content>, and under bad segmentation condition, only<jats:styled-content style="fixed-case">BEE</jats:styled-content>boosted the performance. The application of saliency‐based image processing strategies was verified to be beneficial to object recognition in daily scenes under simulated prosthetic vision. They are hoped to help the development of the image processing module for future retinal prostheses, and thus provide more benefit for the patients.</jats:p>