Salient object detection via reliability‐based depth compactness and depth contrast

Zhou, Yang; Liu, Xiaoqi; Zhang, Yun; Yin, Haibing; Lu, Yu

doi:10.1049/iet-ipr.2019.1495

Cited by 2 publications

(1 citation statement)

References 43 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…They used algorithms that directly detect significant objects for correlation detection, evaluated depth image confidence through depth reliability indicators, reflected main salient features using depth compactness, and enhanced depth contrast through rough background filtering. The outcomes indicated that the method used can perform significance detection well [9]. Wang G et al conducted relevant anti-deception detection based on the characteristics of visual saliency to achieve better anti-deception detection effects in silicone mask attacks.…”

Section: Related Workmentioning

confidence: 99%

The Application of FLICSP Algorithm Based on Multi-Feature Fusion in Image Saliency Detection

Li,

Wu,

Wang

2024

IEEE Access

View full text Add to dashboard Cite

To raise the performance of the current image saliency detection method and promote the extraction effect of salient regions in complex images, the color and texture features fusion is studied on the basis of fast linear iterative clustering with active search to obtain the relevant saliency detection algorithm, and then the fused saliency map is obtained. Algorithm performance is optimized by using a deep prior information-assisted image enhancement model. The outcomes express that compared with other algorithms, the improved algorithm has a smaller mean absolute error, higher precision and F value, lower missed detection rate and false detection rate, and higher peak signal-to-noise ratio and structure similarity index. In the JUDD dataset, the minimum mean absolute error value of the improved algorithm is 0.143, which is 0.34 less than the original algorithm. The improved algorithm in the PASCAL dataset has the highest precision, and F-value, with 0.786 and 0.754 respectively, while the F-value of the pre-improved algorithm is 0.678. In terms of missed detection rate, the improved algorithm is 4.7%, which is 2.5% lower than the previous algorithm; In terms of false detection rates, the pre and post-improvement algorithms have false detection rates of 3.2% and 5.1%, respectively. In the peak signal-to-noise ratio index, the improved algorithm has a maximum value of 39.45 dB, which is 6.82 dB higher than the previous algorithm; Unlike other algorithms, the improved algorithm has the highest similarity index value of 0.892. Research methods can effectively detect the saliency of complex images.

show abstract

Section: Related Workmentioning

confidence: 99%