In computer vision, which visual cues are typically the most informative for scene understanding and object perception in practical systems?

Introduction / Context:
Computer vision systems infer structure and semantics from images or video by exploiting visual cues. Understanding which cues are reliably informative helps in designing feature extractors, sensors, and algorithms for tasks like tracking, segmentation, recognition, and 3D reconstruction.

Given Data / Assumptions:

• Candidates include colour, motion, depth, and texture—well-known cues in vision.
• ‘‘Height and weight’’ are not direct visual cues derived from pixels.
• We are looking for the combination that captures multiple reliable signal sources.

Concept / Approach:

Colour aids segmentation and recognition under consistent illumination; motion provides temporal coherence and can separate foreground from background; depth describes scene geometry; texture conveys surface properties and aids in stereo and shading. Together, these cues complement each other across varied lighting, occlusion, and viewpoint changes. Hence, both (a) and (b) are correct collectively.

Step-by-Step Solution:

Verification / Alternative check:

Standard textbooks and benchmarks (e.g., optical flow for motion, stereo for depth, SIFT/LBP for texture, colour histograms for segmentation) validate the importance of these cues.

Why Other Options Are Wrong:

Only (a) or only (b): Too narrow; both sets are widely used.

Height and weight: Not visual cues readily extracted from raw images.

None: Incorrect because multiple valid cues are listed.

Common Pitfalls:

Assuming colour is always reliable; in practice, illumination changes can degrade it, which is why combining cues is beneficial.

Final Answer:

a and b above