Our ability to see and interact with the world heavily relies on the efficient and precise processing of visual information, consisting of multiple features continuously in flux. Visual perception arises from signals propagating through parallel neuronal pathways wiring the retina to both cortical (primary visual cortex, V1; higher visual areas, HVAs) and subcortical structures (thalamus; superior colliculus, SC), each with distinct functional properties and cellular compositions. However, the precise distribution of behaviorally relevant visual signals across these pathways remains unclear, as probing the entire visual system poses significant challenges. To address this, we combined a visual detection task in mice with mesoscopic optogenetic suppression of cortical activity using patterned-light activation of parvalbumin-positive (PV+) neurons. This allowed us to probe visual stimuli of varying contrast, speed, and spatial frequency, reflecting the diverse visual stimuli encountered in natural environments. Suppressing V1 or HVA activity during low-contrast stimuli and presentation decreased detection rates and increased reaction times, which is consistent with reduced sensory evidence. The effects also persisted when targeting individual HVAs, suggesting widely distributed representations. At high contrast, the impact of suppression depended on stimulus speed and spatial frequency, with pronounced effects observed for slow-moving, high-spatial frequency stimuli and minor effects for fast-moving, low-spatial frequency stimuli. We hypothesized that detecting fast-moving, low-spatial frequency stimuli during cortical silencing reflects signals in SC pathways. Accordingly, chemogenetic silencing of the SC selectively impaired the detection of fast-moving, low-spatial frequency stimuli observed during cortical inhibition, supporting our hypothesis. Our results demonstrate the broad distribution of behaviorally relevant visual information. Cortical and subcortical pathways contribute dynamically to visual perception, with implications for understanding sensory processing and visual conditions such as blindsight. Additionally, these findings offer a model of how redundancy in the visual network supports perceptual robustness.