Abstract :
While many different image properties, such as texture, depth, and homogeneous regions, are suggestive of the structure of an object, on their own, none are sufficient to enable reliable recognition. By fusing these low-level cues and combining with higher-level object shape properties, the method that we describe exploits their complementary evidence. Consequently, this leads to improved robustness and recognition performance.
Keywords :
image texture; object recognition; optimisation; homogeneous regions; image depth; image properties; image texture; low-level cues; multiple cues; object recognition; object shape properties; object structure; optimization; recognition performance; robustness; Humans; Image recognition; Image segmentation; Layout; Object recognition; Permission; Probability density function; Robustness; Shape; Surface texture;
